@sogni-ai/sogni-protocol 1.0.0-alpha.36 → 1.0.0-alpha.39

package/manifests/generation-tools.json

@@ -196,10 +196,6 @@
               "minimum": 2,
               "maximum": 30
             },
-            "smartDuration": {
-              "type": "boolean",
-              "description": "Wan 3 only. Set true to let Wan 3 choose an output length from 2-30 seconds. Do not also set duration. Sogni reserves the 30-second maximum before generation and settles the completed job down to Alibaba's reported output duration."
-            },
             "ratio": {
               "type": "string",
               "enum": [
@@ -248,7 +244,7 @@
                 "wan3.0-video",
                 "wan3.0-spicy-video"
               ],
-              "description": "\"ltx25\" (default): LTX 2.5 with native audio; Fast, HQ, and Pro currently use the release-validated official Distilled INT8 workflow. The Dev checkpoints are not publicly routed until upstream publishes and Sogni validates an official ComfyUI Dev recipe. Video model. \"ltx23\": LTX 2.3 rollback with native audio. \"wan22\": quick simple motion without audio. \"minimax-h3-t2v\": standard 20-step MiniMax H3 text-to-video; \"minimax-h3-t2v-turbo\": the existing 4-step LightX2V Turbo text-to-video; \"minimax-h3-fasth3-t2v-turbo\": the separate FastVideo VSA four-step FastH3 engine, about 2x faster and fixed to Euler/simple. All use native audio, fixed 24fps, 5.17-15.08s, and a 768p-class 32px-grid canvas; use animate_photo for H3 image-conditioned modes. Base and Turbo T2V/I2V/FLF2V prompts use the exact ordered fields integrated_multimodal_description, overall_soundscape, and non_diegetic_music; I2V/FLF2V prepend the official alignment line. \"minimax-h3-r2v\": standard 20-step MiniMax H3 reference-to-video; \"minimax-h3-r2v-turbo\": the dedicated LightX2V 4-step Ref2VA Turbo workflow using Euler/simple and a 960x544 default. FastH3 has no R2V mode. Both R2V selectors accept up to 9 images, 3 videos, and 3 audios (12 files total); at least one visual reference (image or video) is required and audio alone is invalid. Select references with referenceImageIndices/referenceVideoIndices/referenceAudioIndices and address them with the official <Subject N>/<Picture N>/<Video N>/<Audio N> semantics. Seedance quality is selected only by model: use \"seedance2-mini\" for Seedance 2.0 Mini or faster/lower-cost 720p iteration, and use \"seedance2\" for the full Seedance 2.0 model, explicit full-quality requests, 1080p/4K requests, or generated/uploaded storyboard images unless the user explicitly asks for a draft or Mini. Do not use Default Media Quality Fast/HQ/Pro or targetResolution to represent Seedance quality. \"seedance2-5\": Seedance 2.5, the newest Seedance generation — 480p and 720p ONLY (it cannot render 1080p or 4K), 4-30s per clip at a fixed 24 fps, native audio, first-and-last-frame conditioning, and a much larger reference budget than the 2.0 family: up to 30 images, 10 videos, and 10 audios, with up to 50 reference media files total, subject to those per-modality caps. Choose \"seedance2-5\" when the user asks for Seedance 2.5, wants a single continuous Seedance clip longer than 15s (2.5 renders up to 30s in one call instead of being split and stitched), or wants a first-and-last-frame Seedance transition. Keep \"seedance2\" for 1080p/4K requests, which Seedance 2.5 cannot satisfy. Seedance supports multimodal loose reference assets: images (up to 9), videos (up to 3), and audios (up to 3), with no more than 12 asset files total. Use @Image1/@Video1/@Audio1 style references in creative briefs when assigning roles. Assign every useful reference asset a role and prefer positive preservation constraints. If an uploaded video is the source clip to transform, upscale, enhance, restyle, or remaster, use video_to_video with controlMode=\"seedance-v2v\" instead of generate_video referenceVideoIndices. Alibaba HappyHorse 1.1 video models (third-party vendor — requires Premium Spark). Select by mode: \"happyhorse-1.1-t2v\" for text-to-video, \"happyhorse-1.1-i2v\" for image-to-video from one first-frame image, and \"happyhorse-1.1-r2v\" for reference-to-video with up to 9 reference images. Resolutions 720P and 1080P; duration 3-15 seconds at 24 fps; native synchronized audio is always generated (do not set generateAudio or negativePrompt). Supported aspect ratios: 16:9, 9:16, 1:1, 4:3, 3:4, 4:5, 5:4, 9:21, 21:9. HappyHorse 1.1 takes image references only and renders a native synchronized audio track (always on; do not set generateAudio or a negative prompt). Pick the model by mode: happyhorse-1.1-t2v for text-to-video (no reference image), happyhorse-1.1-i2v for image-to-video from a single first frame, and happyhorse-1.1-r2v for reference-to-video with 1 to 9 reference images. For r2v, tag the images in the prompt as [Image 1]…[Image 9] and assign each a clear role. HappyHorse does not accept reference videos or reference audios. \"wan3.0-video\" is Alibaba Wan 3 and \"wan3.0-spicy-video\" is MuleRouter w3.0-video. Both render 2-30s or smart duration at fixed 30 fps with optional native audio, provider prompt expansion, 480p/720p/1080p, adaptive/fixed ratios, first/last frames, and up to 10 image/5 video/5 audio references. Only Alibaba wan3.0-video accepts document/web context and watermark. Frame anchors and loose references are mutually exclusive. Do not send negativePrompt; video references are loose conditioning for a new result, not source-video editing or extension."
+              "description": "\"ltx25\" (default): LTX 2.5 with native audio; Fast, HQ, and Pro currently use the release-validated official Distilled INT8 workflow. The Dev checkpoints are not publicly routed until upstream publishes and Sogni validates an official ComfyUI Dev recipe. Video model. \"ltx23\": LTX 2.3 rollback with native audio. \"wan22\": quick simple motion without audio. \"minimax-h3-t2v\": standard 20-step MiniMax H3 text-to-video; \"minimax-h3-t2v-turbo\": the existing 4-step LightX2V Turbo text-to-video; \"minimax-h3-fasth3-t2v-turbo\": the separate FastVideo VSA four-step FastH3 engine, about 2x faster and fixed to Euler/simple. All use native audio, fixed 24fps, 5.17-15.08s, and a 768p-class 32px-grid canvas; use animate_photo for H3 image-conditioned modes. Base and Turbo T2V/I2V/FLF2V prompts use the exact ordered fields integrated_multimodal_description, overall_soundscape, and non_diegetic_music; I2V/FLF2V prepend the official alignment line. \"minimax-h3-r2v\": standard 20-step MiniMax H3 reference-to-video; \"minimax-h3-r2v-turbo\": the dedicated LightX2V 4-step Ref2VA Turbo workflow using Euler/simple and a 960x544 default. FastH3 has no R2V mode. Both R2V selectors accept up to 9 images, 3 videos, and 3 audios (12 files total); at least one visual reference (image or video) is required and audio alone is invalid. Select references with referenceImageIndices/referenceVideoIndices/referenceAudioIndices and address them with the official <Subject N>/<Picture N>/<Video N>/<Audio N> semantics. Seedance quality is selected only by model: use \"seedance2-mini\" for Seedance 2.0 Mini or faster/lower-cost 720p iteration, and use \"seedance2\" for the full Seedance 2.0 model, explicit full-quality requests, 1080p/4K requests, or generated/uploaded storyboard images unless the user explicitly asks for a draft or Mini. Do not use Default Media Quality Fast/HQ/Pro or targetResolution to represent Seedance quality. \"seedance2-5\": Seedance 2.5, the newest Seedance generation — 480p and 720p ONLY (it cannot render 1080p or 4K), 4-30s per clip at a fixed 24 fps, native audio, first-and-last-frame conditioning, and a much larger reference budget than the 2.0 family: up to 30 images, 10 videos, and 10 audios, with up to 50 reference media files total, subject to those per-modality caps. Choose \"seedance2-5\" when the user asks for Seedance 2.5, wants a single continuous Seedance clip longer than 15s (2.5 renders up to 30s in one call instead of being split and stitched), or wants a first-and-last-frame Seedance transition. Keep \"seedance2\" for 1080p/4K requests, which Seedance 2.5 cannot satisfy. Seedance supports multimodal loose reference assets: images (up to 9), videos (up to 3), and audios (up to 3), with no more than 12 asset files total. Use @Image1/@Video1/@Audio1 style references in creative briefs when assigning roles. Assign every useful reference asset a role and prefer positive preservation constraints. If an uploaded video is the source clip to transform, upscale, enhance, restyle, or remaster, use video_to_video with controlMode=\"seedance-v2v\" instead of generate_video referenceVideoIndices. Alibaba HappyHorse 1.1 video models (third-party vendor — requires Premium Spark). Select by mode: \"happyhorse-1.1-t2v\" for text-to-video, \"happyhorse-1.1-i2v\" for image-to-video from one first-frame image, and \"happyhorse-1.1-r2v\" for reference-to-video with up to 9 reference images. Resolutions 720P and 1080P; duration 3-15 seconds at 24 fps; native synchronized audio is always generated (do not set generateAudio or negativePrompt). Supported aspect ratios: 16:9, 9:16, 1:1, 4:3, 3:4, 4:5, 5:4, 9:21, 21:9. HappyHorse 1.1 takes image references only and renders a native synchronized audio track (always on; do not set generateAudio or a negative prompt). Pick the model by mode: happyhorse-1.1-t2v for text-to-video (no reference image), happyhorse-1.1-i2v for image-to-video from a single first frame, and happyhorse-1.1-r2v for reference-to-video with 1 to 9 reference images. For r2v, tag the images in the prompt as [Image 1]…[Image 9] and assign each a clear role. HappyHorse does not accept reference videos or reference audios. \"wan3.0-video\" is Alibaba Wan 3 and \"wan3.0-spicy-video\" is MuleRouter w3.0-video. Both render 2-30s at fixed 30 fps with optional native audio, provider prompt expansion, 480p/720p/1080p, adaptive/fixed ratios, first/last frames, and up to 10 image/5 video/5 audio references. Only Alibaba wan3.0-video accepts document/web context and watermark. Frame anchors and loose references are mutually exclusive. Do not send negativePrompt; video references are loose conditioning for a new result, not source-video editing or extension."
             },
             "generateAudio": {
               "type": "boolean",
@@ -285,7 +281,15 @@
             },
             "targetResolution": {
               "type": "number",
-              "description": "Short-side video resolution target in pixels. Use when the user asks for a bare named resolution such as \"480p\", \"720p\", \"1080p\", \"2160p\", or \"4K\" without exact pixels or an output orientation. This is resolution only, not a Seedance quality tier: Seedance quality is selected by videoModel (\"seedance2\" vs \"seedance2-mini\" vs \"seedance2-5\"). Seedance 2.0 full supports 4K; Seedance Mini and Seedance 2.5 support 480p/720p only, so never set 1080p or 4K for \"seedance2-5\". Wan 3 supports exactly 480p, 720p, and 1080p. HappyHorse supports only 720p and 1080p. Never set 4K for Wan 3 or HappyHorse. MiniMax H3 renders inside a 1344x768 pixel budget on a 32px grid, so use 768 for H3 and never 1080p or 4K. Do not set targetResolution from Default Media Quality Fast/HQ/Pro. If omitted for Seedance, Wan 3, HappyHorse, or MiniMax H3, the host uses the selected model default. This preserves/inherits the current video shape instead of forcing landscape. Do NOT set width, height, or exact-pixel aspectRatio for bare named resolution requests. If the user says \"720p portrait\", \"720p landscape\", \"4K portrait\", or \"4K landscape\", use exact width/height/aspectRatio instead."
+              "description": "Short-side video resolution target in pixels. Use when the user asks for a bare named resolution such as \"480p\", \"720p\", \"1080p\", \"2160p\", or \"4K\" without exact pixels or an output orientation. This is resolution only, not a Seedance quality tier: Seedance quality is selected by videoModel (\"seedance2\" vs \"seedance2-mini\" vs \"seedance2-5\"). Seedance 2.0 full supports 4K; Seedance Mini and Seedance 2.5 support 480p/720p only, so never set 1080p or 4K for \"seedance2-5\". Wan 3 supports exactly 480p, 720p, and 1080p. HappyHorse supports only 720p and 1080p. Never set 4K for Wan 3 or HappyHorse. MiniMax H3 renders inside a 1344x768 pixel budget on a 32px grid, so use 768 for H3 and never 1080p or 4K. Do not set targetResolution from Default Media Quality Fast/HQ/Pro. If omitted for Seedance, Wan 3, HappyHorse, or MiniMax H3, the host uses the selected model default. This preserves/inherits the current video shape instead of forcing landscape. Do NOT set width, height, or exact-pixel aspectRatio for bare named resolution requests. If the user says \"720p portrait\", \"720p landscape\", \"4K portrait\", or \"4K landscape\", use exact width/height/aspectRatio instead. For MiniMax H3 2K requests keep targetResolution at 768 (or omit it) and set outputScale to 2 instead."
+            },
+            "outputScale": {
+              "type": "integer",
+              "enum": [
+                1,
+                2
+              ],
+              "description": "MiniMax H3 only. 2 delivers 2K output: the clip renders on the normal H3 canvas and is delivered at twice its width and height (1344x768 becomes 2688x1536) with the same length and audio, for +10 Spark per second (+6 at 480p). Set 2 only when the user asks for 2K, 1440p-class or extra-sharp H3 output; leave unset otherwise. Ignored for other models."
             },
             "numberOfVariations": {
               "type": "number",
@@ -309,7 +313,7 @@
                 "type": "string",
                 "minLength": 1
               },
-              "description": "Ordered LoRA IDs to apply to a MiniMax H3 render. Use only when the user explicitly asks for a LoRA or for an effect one of these names describes. Stack up to 8 in one request; order matters because the adapters apply in sequence and do not commute. Keep this array positionally aligned with loraStrengths. The first render with an uncached LoRA takes longer to start while the worker downloads it.\n\nAccepted only when videoModel is one of \"minimax-h3-t2v\", \"minimax-h3-t2v-turbo\", \"minimax-h3-fasth3-t2v-turbo\", \"minimax-h3-r2v\", \"minimax-h3-r2v-turbo\". Every other video model on this tool loads no LoRAs and silently ignores these arrays, so set videoModel to an H3 mode in the same call when the user asks for one.\n\nOne LoRA is published for MiniMax H3 today: h3-realism-people (fal), a realism pass trained on live-action footage of people. It restores skin texture and pores, stray hairs, fabric weave and a fine sensor grain that the base model smooths away, and holds up in close-up. It needs its trigger word: put r34l1sm near the FRONT of the prompt, or the render comes back as ordinary H3 with no error. Exact ranges and any LoRA published since: GET /v1/loras/comfy?modelId=<model>. Do not invent ids."
+              "description": "Ordered LoRA IDs to apply to a MiniMax H3 render. Use only when the user explicitly asks for a LoRA or for an effect one of these names describes. Stack up to 8 in one request; order matters because the adapters apply in sequence and do not commute. Keep this array positionally aligned with loraStrengths. The first render with an uncached LoRA takes longer to start while the worker downloads it.\n\nAccepted only when videoModel is one of \"minimax-h3-t2v\", \"minimax-h3-t2v-turbo\", \"minimax-h3-fasth3-t2v-turbo\", \"minimax-h3-r2v\", \"minimax-h3-r2v-turbo\". Every other video model on this tool loads no LoRAs and silently ignores these arrays, so set videoModel to an H3 mode in the same call when the user asks for one.\n\nFive LoRAs are published for MiniMax H3 today and the set differs by mode, so GET /v1/loras/comfy?modelId=<model> is authoritative for the mode in hand and carries exact ranges, maturity flags, and anything published since. h3-realism-people (fal) is a realism pass trained on live-action footage of people: it restores skin texture and pores, stray hairs, fabric weave and a fine sensor grain that the base model smooths away, and holds up in close-up. It is the only one gated on a trigger word — put r34l1sm near the FRONT of the prompt, or the render comes back as ordinary H3 with no error. h3-vbvr-video-reasoning is a prompt-adherence pass that holds the model to what was asked instead of improvising. h3-natural-face-speech (AdaptiveVision) makes people talking on camera look and sound more natural: cheeks, brows, jaw and lips move together as in real speech, and spoken English comes through clearer; use it for talking-head shots such as vlogs, podcasts, interviews and presenters. h3-better-motion (AdaptiveVision) gives people more natural, consistent body movement — weight shifts, strides, turns and gestures that follow through — for dance, sport, walking and other full-body shots. Both AdaptiveVision LoRAs work best with short, simple prompt sentences and are not validated on reference-to-video. h3-mystic-xxx-v4 is an uncensored adult fine-tune. Do not invent ids."
             },
             "loraStrengths": {
               "type": "array",
@@ -318,7 +322,7 @@
               "items": {
                 "type": "number"
               },
-              "description": "Strength for each LoRA in loras, in the same order. Omitting the array applies 1.0 to every LoRA, which is NOT the catalog default and for h3-realism-people is already at the top of its band, so send explicit values. Video LoRAs are positive-only — unlike the bipolar Krea 2 image sliders, a negative value is not an inverse effect and 0 is off. h3-realism-people takes 0-2 and its catalog default is 0.8; 0.6-1 is the usable band. It also pulls the camera in as it climbs: at 1.5 and above the shot reliably recomposes and the grade darkens, which on an image-conditioned mode can crop the subject out of the frame the user supplied. Raise it above 1 only when the user asks for more, and prefer the default when they supplied a first or last frame."
+              "description": "Strength for each LoRA in loras, in the same order. Omitting the array applies 1.0 to every LoRA, which is NOT the catalog default and for h3-realism-people is already at the top of its band, so send explicit values. Video LoRAs are positive-only — unlike the bipolar Krea 2 image sliders, a negative value is not an inverse effect and 0 is off. h3-realism-people takes 0-2 and its catalog default is 0.8; 0.6-1 is the usable band. It also pulls the camera in as it climbs: at 1.5 and above the shot reliably recomposes and the grade darkens, which on an image-conditioned mode can crop the subject out of the frame the user supplied. Raise it above 1 only when the user asks for more, and prefer the default when they supplied a first or last frame. h3-vbvr-video-reasoning and h3-mystic-xxx-v4 both take 0-1 and do default to 1.0, with usable bands of 0.7-1 and 0.2-1. h3-natural-face-speech and h3-better-motion take 0-1.5 and default to 0.6; their usable band is 0.4-0.8."
             }
           },
           "required": [
@@ -435,8 +439,7 @@
             },
             "voiceSourceIndex": {
               "type": "number",
-              "description": "Index of the uploaded audio file holding the voice to clone (0-based, from the uploaded files list, the same numbering other tools use for audio). Required when model=\"clone\" and ignored otherwise. The clip should be three to thirty seconds of one person speaking cleanly, with no music, no second speaker and no heavy room echo; anything past thirty seconds is trimmed.",
-              "minimum": 0
+              "description": "Which audio holds the voice to clone, using the same numbering every other tool uses for audio: negative indices are uploads (-1 = first/primary upload, -2 = second upload, and so on) and 0-based non-negative indices are audio generated earlier in this conversation. A clip the user just uploaded is -1. Required when model=\"clone\" and ignored otherwise. The clip should be three to thirty seconds of one person speaking cleanly, with no music, no second speaker and no heavy room echo; anything past thirty seconds is trimmed."
             },
             "voiceTranscript": {
               "type": "string",
@@ -712,7 +715,7 @@
       "type": "function",
       "function": {
         "name": "upscale_video",
-        "description": "Upscale an existing video to 1080p or 1440p with FlashVSR video super-resolution. This is promptless, deterministic enhancement, not a generative edit: it keeps every source frame, the exact frame rate, the full aspect ratio, and the original audio, and it never changes content, trims, crops, restyles, or interpolates frames. Use it when the user asks to upscale, enlarge, sharpen, or increase the resolution of an uploaded or previously generated video, or wants an HD, 1080p, 1440p, or 2K copy of it. Do not use it to create, restyle, extend, or edit a video; when the user explicitly asks a generative model such as Seedance to re-render the clip, use video_to_video. The output is at most twice the source size, so 1080p needs a source whose short edge is 540-768px and 1440p needs 720-768px, and the source can be at most about 1344x768 pixels (768x1344 in portrait). It must also be 1-60 fps and at most 100 MB. The server sets the maximum clip length and rejects a source that is too long; never quote a length limit yourself, and if the tool reports that rejection, relay that error to the user. It cannot produce 4K or any size above 1440p; when the user asks for one, say so and offer 1440p. Each upscale costs credits based on the source video's size and length.",
+        "description": "Upscale an existing video to 1080p or 1440p with FlashVSR video super-resolution. This is promptless, deterministic enhancement, not a generative edit: it keeps every source frame, the exact frame rate, the full aspect ratio, and the original audio, and it never changes content, trims, crops, restyles, or interpolates frames. Use it when the user asks to upscale, enlarge, sharpen, or increase the resolution of an uploaded or previously generated video, or wants an HD, 1080p, 1440p, or 2K copy of it. Do not use it to create, restyle, extend, or edit a video; when the user explicitly asks a generative model such as Seedance to re-render the clip, use video_to_video. The output is at most twice the source size, so 1080p needs a source whose short edge is 540-768px and 1440p needs 720-768px, and the source can be at most about 1344x768 pixels (768x1344 in portrait). It must also be 1-60 fps and at most 100 MB. The server sets the maximum clip length and rejects a source that is too long; never quote a length limit yourself, and if the tool reports that rejection, relay that error to the user. It cannot produce 4K or any size above 1440p; when the user asks for one, say so and offer 1440p. Each upscale costs credits based on the source video's size and length. Optional detailPreference (stable or sharper), processingSpeed (stable or faster), and seed tune the result; omit them unless the user asks, and none of them changes the price.",
         "parameters": {
           "type": "object",
           "properties": {
@@ -727,6 +730,28 @@
                 1440
               ],
               "description": "Output resolution of the short edge in pixels: 1440 for 1440p/2K or 1080 for 1080p/Full HD. The long edge follows the source aspect ratio, so portrait videos stay portrait. Default: 1440, or 1080 when the source's short edge is below 720px. Set 1080 when the user asks for 1080p or Full HD."
+            },
+            "detailPreference": {
+              "type": "string",
+              "enum": [
+                "stable",
+                "sharper"
+              ],
+              "description": "Detail preference. stable, the default, gives the most temporally stable result. sharper renders crisper fine detail with a little more risk of shimmer between frames. Set sharper only when the user asks for a sharper, crisper, or more detailed upscale; otherwise omit it. It does not change the price."
+            },
+            "processingSpeed": {
+              "type": "string",
+              "enum": [
+                "stable",
+                "faster"
+              ],
+              "description": "Processing speed. stable, the default, gives the most stable result. faster finishes sooner with slightly less stable detail. Set faster only when the user asks for a quicker upscale; otherwise omit it. It does not change the price."
+            },
+            "seed": {
+              "type": "integer",
+              "minimum": -1,
+              "maximum": 4294967295,
+              "description": "Seed for the upscaler's fine texture, 0 through 4294967295; different seeds give slightly different fine detail. Omit it to keep the repeatable default of 0. Use -1 for a random seed. Set it only when the user gives a seed or asks for a different or random variation of an upscale."
             }
           },
           "required": []
@@ -814,7 +839,7 @@
                 "wan3.0-video",
                 "wan3.0-spicy-video"
               ],
-              "description": "\"ltx25\" (default): LTX 2.5 I2V or first/last-frame video with native audio; Fast, HQ, and Pro currently use the release-validated official Distilled INT8 workflow. The Dev checkpoints are not publicly routed until upstream publishes and Sogni validates an official ComfyUI Dev recipe. Which video model to use. \"ltx23\": LTX 2.3 rollback with native audio. \"wan22\": quick simple motion without audio, up to 10s. \"minimax-h3-i2v\" is standard MiniMax H3 from one first frame; \"minimax-h3-i2v-turbo\" is the existing 4-step LightX2V Turbo engine; \"minimax-h3-fasth3-i2v-turbo\" is the separate FastVideo VSA four-step FastH3 engine, about 2x faster and fixed to Euler/simple. The matching FLF2V selectors provide standard, LightX2V Turbo, and FastH3 Turbo first/last-frame generation; FastH3 has no R2V mode; use frameRole=\"both\" and provide the end frame. H3 generates native audio at fixed 24fps for 5.17-15.08s and has no negative-prompt input. H3 Base and Turbo prompts use the exact three-field contract and the official mode-specific alignment line. Do not set Seedance here; use generate_video with Seedance references. \"wan3.0-video\" is Alibaba Wan 3 and \"wan3.0-spicy-video\" is MuleRouter w3.0-video. Both render 2-30s or smart duration at fixed 30 fps with optional native audio, provider prompt expansion, 480p/720p/1080p, adaptive/fixed ratios, first/last frames, and up to 10 image/5 video/5 audio references. Only Alibaba wan3.0-video accepts document/web context and watermark. Frame anchors and loose references are mutually exclusive. Do not send negativePrompt; video references are loose conditioning for a new result, not source-video editing or extension."
+              "description": "\"ltx25\" (default): LTX 2.5 I2V or first/last-frame video with native audio; Fast, HQ, and Pro currently use the release-validated official Distilled INT8 workflow. The Dev checkpoints are not publicly routed until upstream publishes and Sogni validates an official ComfyUI Dev recipe. Which video model to use. \"ltx23\": LTX 2.3 rollback with native audio. \"wan22\": quick simple motion without audio, up to 10s. \"minimax-h3-i2v\" is standard MiniMax H3 from one first frame; \"minimax-h3-i2v-turbo\" is the existing 4-step LightX2V Turbo engine; \"minimax-h3-fasth3-i2v-turbo\" is the separate FastVideo VSA four-step FastH3 engine, about 2x faster and fixed to Euler/simple. The matching FLF2V selectors provide standard, LightX2V Turbo, and FastH3 Turbo first/last-frame generation; FastH3 has no R2V mode; use frameRole=\"both\" and provide the end frame. H3 generates native audio at fixed 24fps for 5.17-15.08s and has no negative-prompt input. H3 Base and Turbo prompts use the exact three-field contract and the official mode-specific alignment line. Do not set Seedance here; use generate_video with Seedance references. \"wan3.0-video\" is Alibaba Wan 3 and \"wan3.0-spicy-video\" is MuleRouter w3.0-video. Both render 2-30s at fixed 30 fps with optional native audio, provider prompt expansion, 480p/720p/1080p, adaptive/fixed ratios, first/last frames, and up to 10 image/5 video/5 audio references. Only Alibaba wan3.0-video accepts document/web context and watermark. Frame anchors and loose references are mutually exclusive. Do not send negativePrompt; video references are loose conditioning for a new result, not source-video editing or extension."
             },
             "negativePrompt": {
               "type": "string",
@@ -828,10 +853,6 @@
               "type": "number",
               "description": "Video duration in seconds. Default: 5. Use when the user explicitly requests a specific length (e.g., \"make a 10 second video\"). Per-model maximum: ltx25 and ltx23 = 20s, wan22 = 10s (clips longer than this are invalid), wan3.0-video = 30s with a 2s minimum, minimax-h3 = 15.08s with a 5.17s minimum because H3 renders 124-362 frames on a 17-frame grid at a fixed 24 fps. For totals beyond the per-model cap, batch multiple clips via sourceImageIndices instead of requesting a single oversized clip."
             },
-            "smartDuration": {
-              "type": "boolean",
-              "description": "Wan 3 only. Let the model choose 2-30 seconds. Do not also set duration. The 30-second maximum is reserved and the final charge settles down to reported duration."
-            },
             "ratio": {
               "type": "string",
               "enum": [
@@ -850,7 +871,15 @@
             },
             "targetResolution": {
               "type": "number",
-              "description": "Short-side video resolution target in pixels. Use when the user asks for a bare named resolution such as \"480p\", \"720p\", or \"1080p\" without exact pixels or an output orientation. This preserves the source image aspect ratio. Wan 3 supports 480p, 720p, and 1080p; HappyHorse supports only 720p and 1080p. Never set 4K for either. MiniMax H3 renders inside a 1344x768 pixel budget on a 32px grid, so use 768 for H3 and never 1080p or 4K. Do NOT set width, height, or exact-pixel aspectRatio for bare named resolution requests. If the user says \"720p portrait\" or \"720p landscape\", use exact-pixel aspectRatio instead."
+              "description": "Short-side video resolution target in pixels. Use when the user asks for a bare named resolution such as \"480p\", \"720p\", or \"1080p\" without exact pixels or an output orientation. This preserves the source image aspect ratio. Wan 3 supports 480p, 720p, and 1080p; HappyHorse supports only 720p and 1080p. Never set 4K for either. MiniMax H3 renders inside a 1344x768 pixel budget on a 32px grid, so use 768 for H3 and never 1080p or 4K. Do NOT set width, height, or exact-pixel aspectRatio for bare named resolution requests. If the user says \"720p portrait\" or \"720p landscape\", use exact-pixel aspectRatio instead. For MiniMax H3 2K requests keep targetResolution at 768 (or omit it) and set outputScale to 2 instead."
+            },
+            "outputScale": {
+              "type": "integer",
+              "enum": [
+                1,
+                2
+              ],
+              "description": "MiniMax H3 only. 2 delivers 2K output: the clip renders on the normal H3 canvas and is delivered at twice its width and height (1344x768 becomes 2688x1536) with the same length and audio, for +10 Spark per second (+6 at 480p). Set 2 only when the user asks for 2K, 1440p-class or extra-sharp H3 output; leave unset otherwise. Ignored for other models."
             },
             "sourceImageIndex": {
               "type": "number",
@@ -918,7 +947,7 @@
                 "type": "string",
                 "minLength": 1
               },
-              "description": "Ordered LoRA IDs to apply to a MiniMax H3 render. Use only when the user explicitly asks for a LoRA or for an effect one of these names describes. Stack up to 8 in one request; order matters because the adapters apply in sequence and do not commute. Keep this array positionally aligned with loraStrengths. The first render with an uncached LoRA takes longer to start while the worker downloads it.\n\nAccepted only when videoModel is one of \"minimax-h3-i2v\", \"minimax-h3-i2v-turbo\", \"minimax-h3-fasth3-i2v-turbo\", \"minimax-h3-flf2v\", \"minimax-h3-flf2v-turbo\", \"minimax-h3-fasth3-flf2v-turbo\". Every other video model on this tool loads no LoRAs and silently ignores these arrays, so set videoModel to an H3 mode in the same call when the user asks for one.\n\nOne LoRA is published for MiniMax H3 today: h3-realism-people (fal), a realism pass trained on live-action footage of people. It restores skin texture and pores, stray hairs, fabric weave and a fine sensor grain that the base model smooths away, and holds up in close-up. It needs its trigger word: put r34l1sm near the FRONT of the prompt, or the render comes back as ordinary H3 with no error. Exact ranges and any LoRA published since: GET /v1/loras/comfy?modelId=<model>. Do not invent ids."
+              "description": "Ordered LoRA IDs to apply to a MiniMax H3 render. Use only when the user explicitly asks for a LoRA or for an effect one of these names describes. Stack up to 8 in one request; order matters because the adapters apply in sequence and do not commute. Keep this array positionally aligned with loraStrengths. The first render with an uncached LoRA takes longer to start while the worker downloads it.\n\nAccepted only when videoModel is one of \"minimax-h3-i2v\", \"minimax-h3-i2v-turbo\", \"minimax-h3-fasth3-i2v-turbo\", \"minimax-h3-flf2v\", \"minimax-h3-flf2v-turbo\", \"minimax-h3-fasth3-flf2v-turbo\". Every other video model on this tool loads no LoRAs and silently ignores these arrays, so set videoModel to an H3 mode in the same call when the user asks for one.\n\nFive LoRAs are published for MiniMax H3 today and the set differs by mode, so GET /v1/loras/comfy?modelId=<model> is authoritative for the mode in hand and carries exact ranges, maturity flags, and anything published since. h3-realism-people (fal) is a realism pass trained on live-action footage of people: it restores skin texture and pores, stray hairs, fabric weave and a fine sensor grain that the base model smooths away, and holds up in close-up. It is the only one gated on a trigger word — put r34l1sm near the FRONT of the prompt, or the render comes back as ordinary H3 with no error. h3-vbvr-video-reasoning is a prompt-adherence pass that holds the model to what was asked instead of improvising. h3-natural-face-speech (AdaptiveVision) makes people talking on camera look and sound more natural: cheeks, brows, jaw and lips move together as in real speech, and spoken English comes through clearer; use it for talking-head shots such as vlogs, podcasts, interviews and presenters. h3-better-motion (AdaptiveVision) gives people more natural, consistent body movement — weight shifts, strides, turns and gestures that follow through — for dance, sport, walking and other full-body shots. Both AdaptiveVision LoRAs work best with short, simple prompt sentences and are not validated on reference-to-video. h3-mystic-xxx-v4 is an uncensored adult fine-tune. Do not invent ids."
             },
             "loraStrengths": {
               "type": "array",
@@ -927,7 +956,7 @@
               "items": {
                 "type": "number"
               },
-              "description": "Strength for each LoRA in loras, in the same order. Omitting the array applies 1.0 to every LoRA, which is NOT the catalog default and for h3-realism-people is already at the top of its band, so send explicit values. Video LoRAs are positive-only — unlike the bipolar Krea 2 image sliders, a negative value is not an inverse effect and 0 is off. h3-realism-people takes 0-2 and its catalog default is 0.8; 0.6-1 is the usable band. It also pulls the camera in as it climbs: at 1.5 and above the shot reliably recomposes and the grade darkens, which on an image-conditioned mode can crop the subject out of the frame the user supplied. Raise it above 1 only when the user asks for more, and prefer the default when they supplied a first or last frame."
+              "description": "Strength for each LoRA in loras, in the same order. Omitting the array applies 1.0 to every LoRA, which is NOT the catalog default and for h3-realism-people is already at the top of its band, so send explicit values. Video LoRAs are positive-only — unlike the bipolar Krea 2 image sliders, a negative value is not an inverse effect and 0 is off. h3-realism-people takes 0-2 and its catalog default is 0.8; 0.6-1 is the usable band. It also pulls the camera in as it climbs: at 1.5 and above the shot reliably recomposes and the grade darkens, which on an image-conditioned mode can crop the subject out of the frame the user supplied. Raise it above 1 only when the user asks for more, and prefer the default when they supplied a first or last frame. h3-vbvr-video-reasoning and h3-mystic-xxx-v4 both take 0-1 and do default to 1.0, with usable bands of 0.7-1 and 0.2-1. h3-natural-face-speech and h3-better-motion take 0-1.5 and default to 0.6; their usable band is 0.4-0.8."
             }
           },
           "required": [
@@ -1252,7 +1281,7 @@
       "type": "function",
       "function": {
         "name": "sound_to_video",
-        "description": "Generate video synchronized to audio. Use when the user has uploaded an audio file (mp3, wav, m4a, flac) and the audio is the primary sync target, especially uploaded-audio-only workflows. Also use after generate_music (\"turn that song into a video\", \"make a music video from that\"). Auto-detects generated audio from generate_music if no audio file is uploaded. Seedance animate_photo/generate_video can also attach uploaded audio as a loose @Audio reference when an image or video reference anchors the request; use this tool instead when the soundtrack itself should drive the video. If the user provides a reference image, use ltx25-ia2v by default (ltx23-ia2v is rollback); for lip-sync with a face image, use wan-s2v; if no image, use ltx25-a2v by default (ltx23-a2v is rollback). If the user wants dialogue/audio WITHOUT pre-existing audio, use animate_photo instead (LTX 2.5 and LTX 2.3 generate audio natively). Note: Persona voice clips from resolve_personas are NOT used by this tool — for persona voice identity in video, use animate_photo or generate_video with videoModel=\"ltx23\" because LTX 2.5 has no compatible ID-LoRA. LONG AUDIO ON SEEDANCE: Seedance 2.0 and Mini cap each clip at 15s; Seedance 2.5 renders up to 30s in one call, so prefer seedance2-5 for 16-30s audio instead of splitting. When the user uploads audio longer than the per-clip cap of the selected model and Seedance is selected (seedance2, seedance2-mini, or seedance2-5), do NOT clamp to 15s and drop the rest — split the run into multiple sound_to_video calls in the same turn (one per 15s segment, so a 20s audio becomes two clips: audioStart=0 duration=15, then audioStart=15 duration=5) and finish with a single stitch_video call referencing the resulting clip indices in order with audioIndex pointing at the same uploaded audio so the stitched output carries the full original soundtrack. LTX/WAN models accept up to 20s per clip, so single-call is fine for them. Use videoModel=\"wan3.0-video\" when the user explicitly requests Wan 3 audio-driven video. Use videoModel=\"wan3.0-spicy-video\" for Wan 3.0 Enhanced audio-driven video through MuleRouter provider ID w3.0-video; it supports smart duration, adaptive ratios, and provider prompt expansion.",
+        "description": "Generate video synchronized to audio. Use when the user has uploaded an audio file (mp3, wav, m4a, flac) and the audio is the primary sync target, especially uploaded-audio-only workflows. Also use after generate_music (\"turn that song into a video\", \"make a music video from that\"). Auto-detects generated audio from generate_music if no audio file is uploaded. Seedance animate_photo/generate_video can also attach uploaded audio as a loose @Audio reference when an image or video reference anchors the request; use this tool instead when the soundtrack itself should drive the video. If the user provides a reference image, use ltx25-ia2v by default (ltx23-ia2v is rollback); for lip-sync with a face image, use wan-s2v; if no image, use ltx25-a2v by default (ltx23-a2v is rollback). If the user wants dialogue/audio WITHOUT pre-existing audio, use animate_photo instead (LTX 2.5 and LTX 2.3 generate audio natively). Note: Persona voice clips from resolve_personas are NOT used by this tool — for persona voice identity in video, use animate_photo or generate_video with videoModel=\"ltx23\" because LTX 2.5 has no compatible ID-LoRA. LONG AUDIO ON SEEDANCE: Seedance 2.0 and Mini cap each clip at 15s; Seedance 2.5 renders up to 30s in one call, so prefer seedance2-5 for 16-30s audio instead of splitting. When the user uploads audio longer than the per-clip cap of the selected model and Seedance is selected (seedance2, seedance2-mini, or seedance2-5), do NOT clamp to 15s and drop the rest — split the run into multiple sound_to_video calls in the same turn (one per 15s segment, so a 20s audio becomes two clips: audioStart=0 duration=15, then audioStart=15 duration=5) and finish with a single stitch_video call referencing the resulting clip indices in order with audioIndex pointing at the same uploaded audio so the stitched output carries the full original soundtrack. LTX/WAN models accept up to 20s per clip, so single-call is fine for them. Use videoModel=\"wan3.0-video\" when the user explicitly requests Wan 3 audio-driven video. Use videoModel=\"wan3.0-spicy-video\" for Wan 3.0 Enhanced audio-driven video through MuleRouter provider ID w3.0-video; it supports adaptive ratios and provider prompt expansion.",
         "parameters": {
           "type": "object",
           "properties": {
@@ -1287,10 +1316,6 @@
               "minimum": 2,
               "maximum": 30
             },
-            "smartDuration": {
-              "type": "boolean",
-              "description": "Wan 3 only. Let the model choose 2-30 seconds. Do not also set duration. Sogni reserves 30 seconds and settles down to the provider-reported duration."
-            },
             "ratio": {
               "type": "string",
               "enum": [
@@ -1329,7 +1354,7 @@
                 "wan3.0-video",
                 "wan3.0-spicy-video"
               ],
-              "description": "\"ltx25-ia2v\" (default with image) and \"ltx25-a2v\" (default without image): LTX 2.5 image+audio and audio-only modes; Fast, HQ, and Pro currently use the release-validated official Distilled INT8 workflows. The Dev checkpoints are not publicly routed until upstream publishes and Sogni validates official ComfyUI Dev recipes. Video model. \"ltx23-ia2v\" (rollback with image): LTX 2.3 image+audio to video, audio-reactive with a reference image; Fast/HQ use the distilled 8-step worker and Default Media Quality Pro uses the non-distilled dev worker. \"ltx23-a2v\" (rollback without image): LTX 2.3 audio-only to video, no image needed, creates video purely from text prompt + audio with the same quality-tier routing. \"wan-s2v\": WAN 2.2 sound-to-video, best for lip-sync with a face image, fast 4-step. \"seedance2\": full Seedance 2.0 audio-reference video, 4-15s. \"seedance2-mini\": Seedance 2.0 Mini, 720p cap, fastest/lower-cost Seedance option. Seedance quality is selected only by this model value: pick \"seedance2-mini\" for faster/lower-cost drafts or explicit Mini requests, and pick \"seedance2\" for full-quality Seedance or 1080p/4K. Do not infer the Seedance model from Default Media Quality Fast/HQ/Pro. \"seedance2-5\": Seedance 2.5, the newest Seedance generation — 480p and 720p ONLY (it cannot render 1080p or 4K), 4-30s per clip at a fixed 24 fps, native audio, first-and-last-frame conditioning, and a much larger reference budget than the 2.0 family: up to 30 images, 10 videos, and 10 audios, with up to 50 reference media files total, subject to those per-modality caps. Choose \"seedance2-5\" when the user asks for Seedance 2.5, wants a single continuous Seedance clip longer than 15s (2.5 renders up to 30s in one call instead of being split and stitched), or wants a first-and-last-frame Seedance transition. Keep \"seedance2\" for 1080p/4K requests, which Seedance 2.5 cannot satisfy. For Seedance audio-reference prompts, preserve exact spoken dialogue when the user supplied it, and assign @Image1/@Audio1 roles. If the user asks for speech without words, describe the vocal performance without inventing quoted dialogue. Treat lip-sync, voice cloning, and real-human reference behavior as provider-sensitive rather than guaranteed. Omit to auto-select based on whether an image is present. \"wan3.0-video\" is Alibaba Wan 3 and \"wan3.0-spicy-video\" is MuleRouter w3.0-video. Both render 2-30s or smart duration at fixed 30 fps with optional native audio, provider prompt expansion, 480p/720p/1080p, adaptive/fixed ratios, first/last frames, and up to 10 image/5 video/5 audio references. Only Alibaba wan3.0-video accepts document/web context and watermark. Frame anchors and loose references are mutually exclusive. Do not send negativePrompt; video references are loose conditioning for a new result, not source-video editing or extension."
+              "description": "\"ltx25-ia2v\" (default with image) and \"ltx25-a2v\" (default without image): LTX 2.5 image+audio and audio-only modes; Fast, HQ, and Pro currently use the release-validated official Distilled INT8 workflows. The Dev checkpoints are not publicly routed until upstream publishes and Sogni validates official ComfyUI Dev recipes. Video model. \"ltx23-ia2v\" (rollback with image): LTX 2.3 image+audio to video, audio-reactive with a reference image; Fast/HQ use the distilled 8-step worker and Default Media Quality Pro uses the non-distilled dev worker. \"ltx23-a2v\" (rollback without image): LTX 2.3 audio-only to video, no image needed, creates video purely from text prompt + audio with the same quality-tier routing. \"wan-s2v\": WAN 2.2 sound-to-video, best for lip-sync with a face image, fast 4-step. \"seedance2\": full Seedance 2.0 audio-reference video, 4-15s. \"seedance2-mini\": Seedance 2.0 Mini, 720p cap, fastest/lower-cost Seedance option. Seedance quality is selected only by this model value: pick \"seedance2-mini\" for faster/lower-cost drafts or explicit Mini requests, and pick \"seedance2\" for full-quality Seedance or 1080p/4K. Do not infer the Seedance model from Default Media Quality Fast/HQ/Pro. \"seedance2-5\": Seedance 2.5, the newest Seedance generation — 480p and 720p ONLY (it cannot render 1080p or 4K), 4-30s per clip at a fixed 24 fps, native audio, first-and-last-frame conditioning, and a much larger reference budget than the 2.0 family: up to 30 images, 10 videos, and 10 audios, with up to 50 reference media files total, subject to those per-modality caps. Choose \"seedance2-5\" when the user asks for Seedance 2.5, wants a single continuous Seedance clip longer than 15s (2.5 renders up to 30s in one call instead of being split and stitched), or wants a first-and-last-frame Seedance transition. Keep \"seedance2\" for 1080p/4K requests, which Seedance 2.5 cannot satisfy. For Seedance audio-reference prompts, preserve exact spoken dialogue when the user supplied it, and assign @Image1/@Audio1 roles. If the user asks for speech without words, describe the vocal performance without inventing quoted dialogue. Treat lip-sync, voice cloning, and real-human reference behavior as provider-sensitive rather than guaranteed. Omit to auto-select based on whether an image is present. \"wan3.0-video\" is Alibaba Wan 3 and \"wan3.0-spicy-video\" is MuleRouter w3.0-video. Both render 2-30s at fixed 30 fps with optional native audio, provider prompt expansion, 480p/720p/1080p, adaptive/fixed ratios, first/last frames, and up to 10 image/5 video/5 audio references. Only Alibaba wan3.0-video accepts document/web context and watermark. Frame anchors and loose references are mutually exclusive. Do not send negativePrompt; video references are loose conditioning for a new result, not source-video editing or extension."
             },
             "generateAudio": {
               "type": "boolean",

package/manifests/openai-tools.json

@@ -155,10 +155,6 @@
               "minimum": 2,
               "maximum": 30
             },
-            "smartDuration": {
-              "type": "boolean",
-              "description": "Wan 3 only. Set true to let Wan 3 choose an output length from 2-30 seconds. Do not also set duration. Sogni reserves the 30-second maximum before generation and settles the completed job down to Alibaba's reported output duration."
-            },
             "ratio": {
               "type": "string",
               "enum": [
@@ -207,7 +203,7 @@
                 "wan3.0-video",
                 "wan3.0-spicy-video"
               ],
-              "description": "\"ltx25\" (default): LTX 2.5 with native audio; Fast, HQ, and Pro currently use the release-validated official Distilled INT8 workflow. The Dev checkpoints are not publicly routed until upstream publishes and Sogni validates an official ComfyUI Dev recipe. Video model. \"ltx23\": LTX 2.3 rollback with native audio. \"wan22\": quick simple motion without audio. \"minimax-h3-t2v\": standard 20-step MiniMax H3 text-to-video; \"minimax-h3-t2v-turbo\": the existing 4-step LightX2V Turbo text-to-video; \"minimax-h3-fasth3-t2v-turbo\": the separate FastVideo VSA four-step FastH3 engine, about 2x faster and fixed to Euler/simple. All use native audio, fixed 24fps, 5.17-15.08s, and a 768p-class 32px-grid canvas; use animate_photo for H3 image-conditioned modes. Base and Turbo T2V/I2V/FLF2V prompts use the exact ordered fields integrated_multimodal_description, overall_soundscape, and non_diegetic_music; I2V/FLF2V prepend the official alignment line. \"minimax-h3-r2v\": standard 20-step MiniMax H3 reference-to-video; \"minimax-h3-r2v-turbo\": the dedicated LightX2V 4-step Ref2VA Turbo workflow using Euler/simple and a 960x544 default. FastH3 has no R2V mode. Both R2V selectors accept up to 9 images, 3 videos, and 3 audios (12 files total); at least one visual reference (image or video) is required and audio alone is invalid. Select references with referenceImageIndices/referenceVideoIndices/referenceAudioIndices and address them with the official <Subject N>/<Picture N>/<Video N>/<Audio N> semantics. Seedance quality is selected only by model: use \"seedance2-mini\" for Seedance 2.0 Mini or faster/lower-cost 720p iteration, and use \"seedance2\" for the full Seedance 2.0 model, explicit full-quality requests, 1080p/4K requests, or generated/uploaded storyboard images unless the user explicitly asks for a draft or Mini. Do not use Default Media Quality Fast/HQ/Pro or targetResolution to represent Seedance quality. \"seedance2-5\": Seedance 2.5, the newest Seedance generation — 480p and 720p ONLY (it cannot render 1080p or 4K), 4-30s per clip at a fixed 24 fps, native audio, first-and-last-frame conditioning, and a much larger reference budget than the 2.0 family: up to 30 images, 10 videos, and 10 audios, with up to 50 reference media files total, subject to those per-modality caps. Choose \"seedance2-5\" when the user asks for Seedance 2.5, wants a single continuous Seedance clip longer than 15s (2.5 renders up to 30s in one call instead of being split and stitched), or wants a first-and-last-frame Seedance transition. Keep \"seedance2\" for 1080p/4K requests, which Seedance 2.5 cannot satisfy. Seedance supports multimodal loose reference assets: images (up to 9), videos (up to 3), and audios (up to 3), with no more than 12 asset files total. Use @Image1/@Video1/@Audio1 style references in creative briefs when assigning roles. Assign every useful reference asset a role and prefer positive preservation constraints. If an uploaded video is the source clip to transform, upscale, enhance, restyle, or remaster, use video_to_video with controlMode=\"seedance-v2v\" instead of generate_video referenceVideoIndices. Alibaba HappyHorse 1.1 video models (third-party vendor — requires Premium Spark). Select by mode: \"happyhorse-1.1-t2v\" for text-to-video, \"happyhorse-1.1-i2v\" for image-to-video from one first-frame image, and \"happyhorse-1.1-r2v\" for reference-to-video with up to 9 reference images. Resolutions 720P and 1080P; duration 3-15 seconds at 24 fps; native synchronized audio is always generated (do not set generateAudio or negativePrompt). Supported aspect ratios: 16:9, 9:16, 1:1, 4:3, 3:4, 4:5, 5:4, 9:21, 21:9. HappyHorse 1.1 takes image references only and renders a native synchronized audio track (always on; do not set generateAudio or a negative prompt). Pick the model by mode: happyhorse-1.1-t2v for text-to-video (no reference image), happyhorse-1.1-i2v for image-to-video from a single first frame, and happyhorse-1.1-r2v for reference-to-video with 1 to 9 reference images. For r2v, tag the images in the prompt as [Image 1]…[Image 9] and assign each a clear role. HappyHorse does not accept reference videos or reference audios. \"wan3.0-video\" is Alibaba Wan 3 and \"wan3.0-spicy-video\" is MuleRouter w3.0-video. Both render 2-30s or smart duration at fixed 30 fps with optional native audio, provider prompt expansion, 480p/720p/1080p, adaptive/fixed ratios, first/last frames, and up to 10 image/5 video/5 audio references. Only Alibaba wan3.0-video accepts document/web context and watermark. Frame anchors and loose references are mutually exclusive. Do not send negativePrompt; video references are loose conditioning for a new result, not source-video editing or extension."
+              "description": "\"ltx25\" (default): LTX 2.5 with native audio; Fast, HQ, and Pro currently use the release-validated official Distilled INT8 workflow. The Dev checkpoints are not publicly routed until upstream publishes and Sogni validates an official ComfyUI Dev recipe. Video model. \"ltx23\": LTX 2.3 rollback with native audio. \"wan22\": quick simple motion without audio. \"minimax-h3-t2v\": standard 20-step MiniMax H3 text-to-video; \"minimax-h3-t2v-turbo\": the existing 4-step LightX2V Turbo text-to-video; \"minimax-h3-fasth3-t2v-turbo\": the separate FastVideo VSA four-step FastH3 engine, about 2x faster and fixed to Euler/simple. All use native audio, fixed 24fps, 5.17-15.08s, and a 768p-class 32px-grid canvas; use animate_photo for H3 image-conditioned modes. Base and Turbo T2V/I2V/FLF2V prompts use the exact ordered fields integrated_multimodal_description, overall_soundscape, and non_diegetic_music; I2V/FLF2V prepend the official alignment line. \"minimax-h3-r2v\": standard 20-step MiniMax H3 reference-to-video; \"minimax-h3-r2v-turbo\": the dedicated LightX2V 4-step Ref2VA Turbo workflow using Euler/simple and a 960x544 default. FastH3 has no R2V mode. Both R2V selectors accept up to 9 images, 3 videos, and 3 audios (12 files total); at least one visual reference (image or video) is required and audio alone is invalid. Select references with referenceImageIndices/referenceVideoIndices/referenceAudioIndices and address them with the official <Subject N>/<Picture N>/<Video N>/<Audio N> semantics. Seedance quality is selected only by model: use \"seedance2-mini\" for Seedance 2.0 Mini or faster/lower-cost 720p iteration, and use \"seedance2\" for the full Seedance 2.0 model, explicit full-quality requests, 1080p/4K requests, or generated/uploaded storyboard images unless the user explicitly asks for a draft or Mini. Do not use Default Media Quality Fast/HQ/Pro or targetResolution to represent Seedance quality. \"seedance2-5\": Seedance 2.5, the newest Seedance generation — 480p and 720p ONLY (it cannot render 1080p or 4K), 4-30s per clip at a fixed 24 fps, native audio, first-and-last-frame conditioning, and a much larger reference budget than the 2.0 family: up to 30 images, 10 videos, and 10 audios, with up to 50 reference media files total, subject to those per-modality caps. Choose \"seedance2-5\" when the user asks for Seedance 2.5, wants a single continuous Seedance clip longer than 15s (2.5 renders up to 30s in one call instead of being split and stitched), or wants a first-and-last-frame Seedance transition. Keep \"seedance2\" for 1080p/4K requests, which Seedance 2.5 cannot satisfy. Seedance supports multimodal loose reference assets: images (up to 9), videos (up to 3), and audios (up to 3), with no more than 12 asset files total. Use @Image1/@Video1/@Audio1 style references in creative briefs when assigning roles. Assign every useful reference asset a role and prefer positive preservation constraints. If an uploaded video is the source clip to transform, upscale, enhance, restyle, or remaster, use video_to_video with controlMode=\"seedance-v2v\" instead of generate_video referenceVideoIndices. Alibaba HappyHorse 1.1 video models (third-party vendor — requires Premium Spark). Select by mode: \"happyhorse-1.1-t2v\" for text-to-video, \"happyhorse-1.1-i2v\" for image-to-video from one first-frame image, and \"happyhorse-1.1-r2v\" for reference-to-video with up to 9 reference images. Resolutions 720P and 1080P; duration 3-15 seconds at 24 fps; native synchronized audio is always generated (do not set generateAudio or negativePrompt). Supported aspect ratios: 16:9, 9:16, 1:1, 4:3, 3:4, 4:5, 5:4, 9:21, 21:9. HappyHorse 1.1 takes image references only and renders a native synchronized audio track (always on; do not set generateAudio or a negative prompt). Pick the model by mode: happyhorse-1.1-t2v for text-to-video (no reference image), happyhorse-1.1-i2v for image-to-video from a single first frame, and happyhorse-1.1-r2v for reference-to-video with 1 to 9 reference images. For r2v, tag the images in the prompt as [Image 1]…[Image 9] and assign each a clear role. HappyHorse does not accept reference videos or reference audios. \"wan3.0-video\" is Alibaba Wan 3 and \"wan3.0-spicy-video\" is MuleRouter w3.0-video. Both render 2-30s at fixed 30 fps with optional native audio, provider prompt expansion, 480p/720p/1080p, adaptive/fixed ratios, first/last frames, and up to 10 image/5 video/5 audio references. Only Alibaba wan3.0-video accepts document/web context and watermark. Frame anchors and loose references are mutually exclusive. Do not send negativePrompt; video references are loose conditioning for a new result, not source-video editing or extension."
             },
             "generateAudio": {
               "type": "boolean",
@@ -244,7 +240,15 @@
             },
             "targetResolution": {
               "type": "number",
-              "description": "Short-side video resolution target in pixels. Use when the user asks for a bare named resolution such as \"480p\", \"720p\", \"1080p\", \"2160p\", or \"4K\" without exact pixels or an output orientation. This is resolution only, not a Seedance quality tier: Seedance quality is selected by videoModel (\"seedance2\" vs \"seedance2-mini\" vs \"seedance2-5\"). Seedance 2.0 full supports 4K; Seedance Mini and Seedance 2.5 support 480p/720p only, so never set 1080p or 4K for \"seedance2-5\". Wan 3 supports exactly 480p, 720p, and 1080p. HappyHorse supports only 720p and 1080p. Never set 4K for Wan 3 or HappyHorse. MiniMax H3 renders inside a 1344x768 pixel budget on a 32px grid, so use 768 for H3 and never 1080p or 4K. Do not set targetResolution from Default Media Quality Fast/HQ/Pro. If omitted for Seedance, Wan 3, HappyHorse, or MiniMax H3, the host uses the selected model default. This preserves/inherits the current video shape instead of forcing landscape. Do NOT set width, height, or exact-pixel aspectRatio for bare named resolution requests. If the user says \"720p portrait\", \"720p landscape\", \"4K portrait\", or \"4K landscape\", use exact width/height/aspectRatio instead."
+              "description": "Short-side video resolution target in pixels. Use when the user asks for a bare named resolution such as \"480p\", \"720p\", \"1080p\", \"2160p\", or \"4K\" without exact pixels or an output orientation. This is resolution only, not a Seedance quality tier: Seedance quality is selected by videoModel (\"seedance2\" vs \"seedance2-mini\" vs \"seedance2-5\"). Seedance 2.0 full supports 4K; Seedance Mini and Seedance 2.5 support 480p/720p only, so never set 1080p or 4K for \"seedance2-5\". Wan 3 supports exactly 480p, 720p, and 1080p. HappyHorse supports only 720p and 1080p. Never set 4K for Wan 3 or HappyHorse. MiniMax H3 renders inside a 1344x768 pixel budget on a 32px grid, so use 768 for H3 and never 1080p or 4K. Do not set targetResolution from Default Media Quality Fast/HQ/Pro. If omitted for Seedance, Wan 3, HappyHorse, or MiniMax H3, the host uses the selected model default. This preserves/inherits the current video shape instead of forcing landscape. Do NOT set width, height, or exact-pixel aspectRatio for bare named resolution requests. If the user says \"720p portrait\", \"720p landscape\", \"4K portrait\", or \"4K landscape\", use exact width/height/aspectRatio instead. For MiniMax H3 2K requests keep targetResolution at 768 (or omit it) and set outputScale to 2 instead."
+            },
+            "outputScale": {
+              "type": "integer",
+              "enum": [
+                1,
+                2
+              ],
+              "description": "MiniMax H3 only. 2 delivers 2K output: the clip renders on the normal H3 canvas and is delivered at twice its width and height (1344x768 becomes 2688x1536) with the same length and audio, for +10 Spark per second (+6 at 480p). Set 2 only when the user asks for 2K, 1440p-class or extra-sharp H3 output; leave unset otherwise. Ignored for other models."
             },
             "numberOfVariations": {
               "type": "number",
@@ -268,7 +272,7 @@
                 "type": "string",
                 "minLength": 1
               },
-              "description": "Ordered LoRA IDs to apply to a MiniMax H3 render. Use only when the user explicitly asks for a LoRA or for an effect one of these names describes. Stack up to 8 in one request; order matters because the adapters apply in sequence and do not commute. Keep this array positionally aligned with loraStrengths. The first render with an uncached LoRA takes longer to start while the worker downloads it.\n\nAccepted only when videoModel is one of \"minimax-h3-t2v\", \"minimax-h3-t2v-turbo\", \"minimax-h3-fasth3-t2v-turbo\", \"minimax-h3-r2v\", \"minimax-h3-r2v-turbo\". Every other video model on this tool loads no LoRAs and silently ignores these arrays, so set videoModel to an H3 mode in the same call when the user asks for one.\n\nOne LoRA is published for MiniMax H3 today: h3-realism-people (fal), a realism pass trained on live-action footage of people. It restores skin texture and pores, stray hairs, fabric weave and a fine sensor grain that the base model smooths away, and holds up in close-up. It needs its trigger word: put r34l1sm near the FRONT of the prompt, or the render comes back as ordinary H3 with no error. Exact ranges and any LoRA published since: GET /v1/loras/comfy?modelId=<model>. Do not invent ids."
+              "description": "Ordered LoRA IDs to apply to a MiniMax H3 render. Use only when the user explicitly asks for a LoRA or for an effect one of these names describes. Stack up to 8 in one request; order matters because the adapters apply in sequence and do not commute. Keep this array positionally aligned with loraStrengths. The first render with an uncached LoRA takes longer to start while the worker downloads it.\n\nAccepted only when videoModel is one of \"minimax-h3-t2v\", \"minimax-h3-t2v-turbo\", \"minimax-h3-fasth3-t2v-turbo\", \"minimax-h3-r2v\", \"minimax-h3-r2v-turbo\". Every other video model on this tool loads no LoRAs and silently ignores these arrays, so set videoModel to an H3 mode in the same call when the user asks for one.\n\nFive LoRAs are published for MiniMax H3 today and the set differs by mode, so GET /v1/loras/comfy?modelId=<model> is authoritative for the mode in hand and carries exact ranges, maturity flags, and anything published since. h3-realism-people (fal) is a realism pass trained on live-action footage of people: it restores skin texture and pores, stray hairs, fabric weave and a fine sensor grain that the base model smooths away, and holds up in close-up. It is the only one gated on a trigger word — put r34l1sm near the FRONT of the prompt, or the render comes back as ordinary H3 with no error. h3-vbvr-video-reasoning is a prompt-adherence pass that holds the model to what was asked instead of improvising. h3-natural-face-speech (AdaptiveVision) makes people talking on camera look and sound more natural: cheeks, brows, jaw and lips move together as in real speech, and spoken English comes through clearer; use it for talking-head shots such as vlogs, podcasts, interviews and presenters. h3-better-motion (AdaptiveVision) gives people more natural, consistent body movement — weight shifts, strides, turns and gestures that follow through — for dance, sport, walking and other full-body shots. Both AdaptiveVision LoRAs work best with short, simple prompt sentences and are not validated on reference-to-video. h3-mystic-xxx-v4 is an uncensored adult fine-tune. Do not invent ids."
             },
             "loraStrengths": {
               "type": "array",
@@ -277,7 +281,7 @@
               "items": {
                 "type": "number"
               },
-              "description": "Strength for each LoRA in loras, in the same order. Omitting the array applies 1.0 to every LoRA, which is NOT the catalog default and for h3-realism-people is already at the top of its band, so send explicit values. Video LoRAs are positive-only — unlike the bipolar Krea 2 image sliders, a negative value is not an inverse effect and 0 is off. h3-realism-people takes 0-2 and its catalog default is 0.8; 0.6-1 is the usable band. It also pulls the camera in as it climbs: at 1.5 and above the shot reliably recomposes and the grade darkens, which on an image-conditioned mode can crop the subject out of the frame the user supplied. Raise it above 1 only when the user asks for more, and prefer the default when they supplied a first or last frame."
+              "description": "Strength for each LoRA in loras, in the same order. Omitting the array applies 1.0 to every LoRA, which is NOT the catalog default and for h3-realism-people is already at the top of its band, so send explicit values. Video LoRAs are positive-only — unlike the bipolar Krea 2 image sliders, a negative value is not an inverse effect and 0 is off. h3-realism-people takes 0-2 and its catalog default is 0.8; 0.6-1 is the usable band. It also pulls the camera in as it climbs: at 1.5 and above the shot reliably recomposes and the grade darkens, which on an image-conditioned mode can crop the subject out of the frame the user supplied. Raise it above 1 only when the user asks for more, and prefer the default when they supplied a first or last frame. h3-vbvr-video-reasoning and h3-mystic-xxx-v4 both take 0-1 and do default to 1.0, with usable bands of 0.7-1 and 0.2-1. h3-natural-face-speech and h3-better-motion take 0-1.5 and default to 0.6; their usable band is 0.4-0.8."
             }
           },
           "required": [
@@ -394,8 +398,7 @@
             },
             "voiceSourceIndex": {
               "type": "number",
-              "description": "Index of the uploaded audio file holding the voice to clone (0-based, from the uploaded files list, the same numbering other tools use for audio). Required when model=\"clone\" and ignored otherwise. The clip should be three to thirty seconds of one person speaking cleanly, with no music, no second speaker and no heavy room echo; anything past thirty seconds is trimmed.",
-              "minimum": 0
+              "description": "Which audio holds the voice to clone, using the same numbering every other tool uses for audio: negative indices are uploads (-1 = first/primary upload, -2 = second upload, and so on) and 0-based non-negative indices are audio generated earlier in this conversation. A clip the user just uploaded is -1. Required when model=\"clone\" and ignored otherwise. The clip should be three to thirty seconds of one person speaking cleanly, with no music, no second speaker and no heavy room echo; anything past thirty seconds is trimmed."
             },
             "voiceTranscript": {
               "type": "string",
@@ -652,7 +655,7 @@
       "type": "function",
       "function": {
         "name": "upscale_video",
-        "description": "Upscale an existing video to 1080p or 1440p with FlashVSR video super-resolution. This is promptless, deterministic enhancement, not a generative edit: it keeps every source frame, the exact frame rate, the full aspect ratio, and the original audio, and it never changes content, trims, crops, restyles, or interpolates frames. Use it when the user asks to upscale, enlarge, sharpen, or increase the resolution of an uploaded or previously generated video, or wants an HD, 1080p, 1440p, or 2K copy of it. Do not use it to create, restyle, extend, or edit a video; when the user explicitly asks a generative model such as Seedance to re-render the clip, use video_to_video. The output is at most twice the source size, so 1080p needs a source whose short edge is 540-768px and 1440p needs 720-768px, and the source can be at most about 1344x768 pixels (768x1344 in portrait). It must also be 1-60 fps and at most 100 MB. The server sets the maximum clip length and rejects a source that is too long; never quote a length limit yourself, and if the tool reports that rejection, relay that error to the user. It cannot produce 4K or any size above 1440p; when the user asks for one, say so and offer 1440p. Each upscale costs credits based on the source video's size and length.",
+        "description": "Upscale an existing video to 1080p or 1440p with FlashVSR video super-resolution. This is promptless, deterministic enhancement, not a generative edit: it keeps every source frame, the exact frame rate, the full aspect ratio, and the original audio, and it never changes content, trims, crops, restyles, or interpolates frames. Use it when the user asks to upscale, enlarge, sharpen, or increase the resolution of an uploaded or previously generated video, or wants an HD, 1080p, 1440p, or 2K copy of it. Do not use it to create, restyle, extend, or edit a video; when the user explicitly asks a generative model such as Seedance to re-render the clip, use video_to_video. The output is at most twice the source size, so 1080p needs a source whose short edge is 540-768px and 1440p needs 720-768px, and the source can be at most about 1344x768 pixels (768x1344 in portrait). It must also be 1-60 fps and at most 100 MB. The server sets the maximum clip length and rejects a source that is too long; never quote a length limit yourself, and if the tool reports that rejection, relay that error to the user. It cannot produce 4K or any size above 1440p; when the user asks for one, say so and offer 1440p. Each upscale costs credits based on the source video's size and length. Optional detailPreference (stable or sharper), processingSpeed (stable or faster), and seed tune the result; omit them unless the user asks, and none of them changes the price.",
         "parameters": {
           "type": "object",
           "properties": {
@@ -667,6 +670,28 @@
                 1440
               ],
               "description": "Output resolution of the short edge in pixels: 1440 for 1440p/2K or 1080 for 1080p/Full HD. The long edge follows the source aspect ratio, so portrait videos stay portrait. Default: 1440, or 1080 when the source's short edge is below 720px. Set 1080 when the user asks for 1080p or Full HD."
+            },
+            "detailPreference": {
+              "type": "string",
+              "enum": [
+                "stable",
+                "sharper"
+              ],
+              "description": "Detail preference. stable, the default, gives the most temporally stable result. sharper renders crisper fine detail with a little more risk of shimmer between frames. Set sharper only when the user asks for a sharper, crisper, or more detailed upscale; otherwise omit it. It does not change the price."
+            },
+            "processingSpeed": {
+              "type": "string",
+              "enum": [
+                "stable",
+                "faster"
+              ],
+              "description": "Processing speed. stable, the default, gives the most stable result. faster finishes sooner with slightly less stable detail. Set faster only when the user asks for a quicker upscale; otherwise omit it. It does not change the price."
+            },
+            "seed": {
+              "type": "integer",
+              "minimum": -1,
+              "maximum": 4294967295,
+              "description": "Seed for the upscaler's fine texture, 0 through 4294967295; different seeds give slightly different fine detail. Omit it to keep the repeatable default of 0. Use -1 for a random seed. Set it only when the user gives a seed or asks for a different or random variation of an upscale."
             }
           },
           "required": []
@@ -754,7 +779,7 @@
                 "wan3.0-video",
                 "wan3.0-spicy-video"
               ],
-              "description": "\"ltx25\" (default): LTX 2.5 I2V or first/last-frame video with native audio; Fast, HQ, and Pro currently use the release-validated official Distilled INT8 workflow. The Dev checkpoints are not publicly routed until upstream publishes and Sogni validates an official ComfyUI Dev recipe. Which video model to use. \"ltx23\": LTX 2.3 rollback with native audio. \"wan22\": quick simple motion without audio, up to 10s. \"minimax-h3-i2v\" is standard MiniMax H3 from one first frame; \"minimax-h3-i2v-turbo\" is the existing 4-step LightX2V Turbo engine; \"minimax-h3-fasth3-i2v-turbo\" is the separate FastVideo VSA four-step FastH3 engine, about 2x faster and fixed to Euler/simple. The matching FLF2V selectors provide standard, LightX2V Turbo, and FastH3 Turbo first/last-frame generation; FastH3 has no R2V mode; use frameRole=\"both\" and provide the end frame. H3 generates native audio at fixed 24fps for 5.17-15.08s and has no negative-prompt input. H3 Base and Turbo prompts use the exact three-field contract and the official mode-specific alignment line. Do not set Seedance here; use generate_video with Seedance references. \"wan3.0-video\" is Alibaba Wan 3 and \"wan3.0-spicy-video\" is MuleRouter w3.0-video. Both render 2-30s or smart duration at fixed 30 fps with optional native audio, provider prompt expansion, 480p/720p/1080p, adaptive/fixed ratios, first/last frames, and up to 10 image/5 video/5 audio references. Only Alibaba wan3.0-video accepts document/web context and watermark. Frame anchors and loose references are mutually exclusive. Do not send negativePrompt; video references are loose conditioning for a new result, not source-video editing or extension."
+              "description": "\"ltx25\" (default): LTX 2.5 I2V or first/last-frame video with native audio; Fast, HQ, and Pro currently use the release-validated official Distilled INT8 workflow. The Dev checkpoints are not publicly routed until upstream publishes and Sogni validates an official ComfyUI Dev recipe. Which video model to use. \"ltx23\": LTX 2.3 rollback with native audio. \"wan22\": quick simple motion without audio, up to 10s. \"minimax-h3-i2v\" is standard MiniMax H3 from one first frame; \"minimax-h3-i2v-turbo\" is the existing 4-step LightX2V Turbo engine; \"minimax-h3-fasth3-i2v-turbo\" is the separate FastVideo VSA four-step FastH3 engine, about 2x faster and fixed to Euler/simple. The matching FLF2V selectors provide standard, LightX2V Turbo, and FastH3 Turbo first/last-frame generation; FastH3 has no R2V mode; use frameRole=\"both\" and provide the end frame. H3 generates native audio at fixed 24fps for 5.17-15.08s and has no negative-prompt input. H3 Base and Turbo prompts use the exact three-field contract and the official mode-specific alignment line. Do not set Seedance here; use generate_video with Seedance references. \"wan3.0-video\" is Alibaba Wan 3 and \"wan3.0-spicy-video\" is MuleRouter w3.0-video. Both render 2-30s at fixed 30 fps with optional native audio, provider prompt expansion, 480p/720p/1080p, adaptive/fixed ratios, first/last frames, and up to 10 image/5 video/5 audio references. Only Alibaba wan3.0-video accepts document/web context and watermark. Frame anchors and loose references are mutually exclusive. Do not send negativePrompt; video references are loose conditioning for a new result, not source-video editing or extension."
             },
             "negativePrompt": {
               "type": "string",
@@ -768,10 +793,6 @@
               "type": "number",
               "description": "Video duration in seconds. Default: 5. Use when the user explicitly requests a specific length (e.g., \"make a 10 second video\"). Per-model maximum: ltx25 and ltx23 = 20s, wan22 = 10s (clips longer than this are invalid), wan3.0-video = 30s with a 2s minimum, minimax-h3 = 15.08s with a 5.17s minimum because H3 renders 124-362 frames on a 17-frame grid at a fixed 24 fps. For totals beyond the per-model cap, batch multiple clips via sourceImageIndices instead of requesting a single oversized clip."
             },
-            "smartDuration": {
-              "type": "boolean",
-              "description": "Wan 3 only. Let the model choose 2-30 seconds. Do not also set duration. The 30-second maximum is reserved and the final charge settles down to reported duration."
-            },
             "ratio": {
               "type": "string",
               "enum": [
@@ -790,7 +811,15 @@
             },
             "targetResolution": {
               "type": "number",
-              "description": "Short-side video resolution target in pixels. Use when the user asks for a bare named resolution such as \"480p\", \"720p\", or \"1080p\" without exact pixels or an output orientation. This preserves the source image aspect ratio. Wan 3 supports 480p, 720p, and 1080p; HappyHorse supports only 720p and 1080p. Never set 4K for either. MiniMax H3 renders inside a 1344x768 pixel budget on a 32px grid, so use 768 for H3 and never 1080p or 4K. Do NOT set width, height, or exact-pixel aspectRatio for bare named resolution requests. If the user says \"720p portrait\" or \"720p landscape\", use exact-pixel aspectRatio instead."
+              "description": "Short-side video resolution target in pixels. Use when the user asks for a bare named resolution such as \"480p\", \"720p\", or \"1080p\" without exact pixels or an output orientation. This preserves the source image aspect ratio. Wan 3 supports 480p, 720p, and 1080p; HappyHorse supports only 720p and 1080p. Never set 4K for either. MiniMax H3 renders inside a 1344x768 pixel budget on a 32px grid, so use 768 for H3 and never 1080p or 4K. Do NOT set width, height, or exact-pixel aspectRatio for bare named resolution requests. If the user says \"720p portrait\" or \"720p landscape\", use exact-pixel aspectRatio instead. For MiniMax H3 2K requests keep targetResolution at 768 (or omit it) and set outputScale to 2 instead."
+            },
+            "outputScale": {
+              "type": "integer",
+              "enum": [
+                1,
+                2
+              ],
+              "description": "MiniMax H3 only. 2 delivers 2K output: the clip renders on the normal H3 canvas and is delivered at twice its width and height (1344x768 becomes 2688x1536) with the same length and audio, for +10 Spark per second (+6 at 480p). Set 2 only when the user asks for 2K, 1440p-class or extra-sharp H3 output; leave unset otherwise. Ignored for other models."
             },
             "sourceImageIndex": {
               "type": "number",
@@ -858,7 +887,7 @@
                 "type": "string",
                 "minLength": 1
               },
-              "description": "Ordered LoRA IDs to apply to a MiniMax H3 render. Use only when the user explicitly asks for a LoRA or for an effect one of these names describes. Stack up to 8 in one request; order matters because the adapters apply in sequence and do not commute. Keep this array positionally aligned with loraStrengths. The first render with an uncached LoRA takes longer to start while the worker downloads it.\n\nAccepted only when videoModel is one of \"minimax-h3-i2v\", \"minimax-h3-i2v-turbo\", \"minimax-h3-fasth3-i2v-turbo\", \"minimax-h3-flf2v\", \"minimax-h3-flf2v-turbo\", \"minimax-h3-fasth3-flf2v-turbo\". Every other video model on this tool loads no LoRAs and silently ignores these arrays, so set videoModel to an H3 mode in the same call when the user asks for one.\n\nOne LoRA is published for MiniMax H3 today: h3-realism-people (fal), a realism pass trained on live-action footage of people. It restores skin texture and pores, stray hairs, fabric weave and a fine sensor grain that the base model smooths away, and holds up in close-up. It needs its trigger word: put r34l1sm near the FRONT of the prompt, or the render comes back as ordinary H3 with no error. Exact ranges and any LoRA published since: GET /v1/loras/comfy?modelId=<model>. Do not invent ids."
+              "description": "Ordered LoRA IDs to apply to a MiniMax H3 render. Use only when the user explicitly asks for a LoRA or for an effect one of these names describes. Stack up to 8 in one request; order matters because the adapters apply in sequence and do not commute. Keep this array positionally aligned with loraStrengths. The first render with an uncached LoRA takes longer to start while the worker downloads it.\n\nAccepted only when videoModel is one of \"minimax-h3-i2v\", \"minimax-h3-i2v-turbo\", \"minimax-h3-fasth3-i2v-turbo\", \"minimax-h3-flf2v\", \"minimax-h3-flf2v-turbo\", \"minimax-h3-fasth3-flf2v-turbo\". Every other video model on this tool loads no LoRAs and silently ignores these arrays, so set videoModel to an H3 mode in the same call when the user asks for one.\n\nFive LoRAs are published for MiniMax H3 today and the set differs by mode, so GET /v1/loras/comfy?modelId=<model> is authoritative for the mode in hand and carries exact ranges, maturity flags, and anything published since. h3-realism-people (fal) is a realism pass trained on live-action footage of people: it restores skin texture and pores, stray hairs, fabric weave and a fine sensor grain that the base model smooths away, and holds up in close-up. It is the only one gated on a trigger word — put r34l1sm near the FRONT of the prompt, or the render comes back as ordinary H3 with no error. h3-vbvr-video-reasoning is a prompt-adherence pass that holds the model to what was asked instead of improvising. h3-natural-face-speech (AdaptiveVision) makes people talking on camera look and sound more natural: cheeks, brows, jaw and lips move together as in real speech, and spoken English comes through clearer; use it for talking-head shots such as vlogs, podcasts, interviews and presenters. h3-better-motion (AdaptiveVision) gives people more natural, consistent body movement — weight shifts, strides, turns and gestures that follow through — for dance, sport, walking and other full-body shots. Both AdaptiveVision LoRAs work best with short, simple prompt sentences and are not validated on reference-to-video. h3-mystic-xxx-v4 is an uncensored adult fine-tune. Do not invent ids."
             },
             "loraStrengths": {
               "type": "array",
@@ -867,7 +896,7 @@
               "items": {
                 "type": "number"
               },
-              "description": "Strength for each LoRA in loras, in the same order. Omitting the array applies 1.0 to every LoRA, which is NOT the catalog default and for h3-realism-people is already at the top of its band, so send explicit values. Video LoRAs are positive-only — unlike the bipolar Krea 2 image sliders, a negative value is not an inverse effect and 0 is off. h3-realism-people takes 0-2 and its catalog default is 0.8; 0.6-1 is the usable band. It also pulls the camera in as it climbs: at 1.5 and above the shot reliably recomposes and the grade darkens, which on an image-conditioned mode can crop the subject out of the frame the user supplied. Raise it above 1 only when the user asks for more, and prefer the default when they supplied a first or last frame."
+              "description": "Strength for each LoRA in loras, in the same order. Omitting the array applies 1.0 to every LoRA, which is NOT the catalog default and for h3-realism-people is already at the top of its band, so send explicit values. Video LoRAs are positive-only — unlike the bipolar Krea 2 image sliders, a negative value is not an inverse effect and 0 is off. h3-realism-people takes 0-2 and its catalog default is 0.8; 0.6-1 is the usable band. It also pulls the camera in as it climbs: at 1.5 and above the shot reliably recomposes and the grade darkens, which on an image-conditioned mode can crop the subject out of the frame the user supplied. Raise it above 1 only when the user asks for more, and prefer the default when they supplied a first or last frame. h3-vbvr-video-reasoning and h3-mystic-xxx-v4 both take 0-1 and do default to 1.0, with usable bands of 0.7-1 and 0.2-1. h3-natural-face-speech and h3-better-motion take 0-1.5 and default to 0.6; their usable band is 0.4-0.8."
             }
           },
           "required": [
@@ -1192,7 +1221,7 @@
       "type": "function",
       "function": {
         "name": "sound_to_video",
-        "description": "Generate video synchronized to audio. Use when the user has uploaded an audio file (mp3, wav, m4a, flac) and the audio is the primary sync target, especially uploaded-audio-only workflows. Also use after generate_music (\"turn that song into a video\", \"make a music video from that\"). Auto-detects generated audio from generate_music if no audio file is uploaded. Seedance animate_photo/generate_video can also attach uploaded audio as a loose @Audio reference when an image or video reference anchors the request; use this tool instead when the soundtrack itself should drive the video. If the user provides a reference image, use ltx25-ia2v by default (ltx23-ia2v is rollback); for lip-sync with a face image, use wan-s2v; if no image, use ltx25-a2v by default (ltx23-a2v is rollback). If the user wants dialogue/audio WITHOUT pre-existing audio, use animate_photo instead (LTX 2.5 and LTX 2.3 generate audio natively). Note: Persona voice clips from resolve_personas are NOT used by this tool — for persona voice identity in video, use animate_photo or generate_video with videoModel=\"ltx23\" because LTX 2.5 has no compatible ID-LoRA. LONG AUDIO ON SEEDANCE: Seedance 2.0 and Mini cap each clip at 15s; Seedance 2.5 renders up to 30s in one call, so prefer seedance2-5 for 16-30s audio instead of splitting. When the user uploads audio longer than the per-clip cap of the selected model and Seedance is selected (seedance2, seedance2-mini, or seedance2-5), do NOT clamp to 15s and drop the rest — split the run into multiple sound_to_video calls in the same turn (one per 15s segment, so a 20s audio becomes two clips: audioStart=0 duration=15, then audioStart=15 duration=5) and finish with a single stitch_video call referencing the resulting clip indices in order with audioIndex pointing at the same uploaded audio so the stitched output carries the full original soundtrack. LTX/WAN models accept up to 20s per clip, so single-call is fine for them. Use videoModel=\"wan3.0-video\" when the user explicitly requests Wan 3 audio-driven video. Use videoModel=\"wan3.0-spicy-video\" for Wan 3.0 Enhanced audio-driven video through MuleRouter provider ID w3.0-video; it supports smart duration, adaptive ratios, and provider prompt expansion.",
+        "description": "Generate video synchronized to audio. Use when the user has uploaded an audio file (mp3, wav, m4a, flac) and the audio is the primary sync target, especially uploaded-audio-only workflows. Also use after generate_music (\"turn that song into a video\", \"make a music video from that\"). Auto-detects generated audio from generate_music if no audio file is uploaded. Seedance animate_photo/generate_video can also attach uploaded audio as a loose @Audio reference when an image or video reference anchors the request; use this tool instead when the soundtrack itself should drive the video. If the user provides a reference image, use ltx25-ia2v by default (ltx23-ia2v is rollback); for lip-sync with a face image, use wan-s2v; if no image, use ltx25-a2v by default (ltx23-a2v is rollback). If the user wants dialogue/audio WITHOUT pre-existing audio, use animate_photo instead (LTX 2.5 and LTX 2.3 generate audio natively). Note: Persona voice clips from resolve_personas are NOT used by this tool — for persona voice identity in video, use animate_photo or generate_video with videoModel=\"ltx23\" because LTX 2.5 has no compatible ID-LoRA. LONG AUDIO ON SEEDANCE: Seedance 2.0 and Mini cap each clip at 15s; Seedance 2.5 renders up to 30s in one call, so prefer seedance2-5 for 16-30s audio instead of splitting. When the user uploads audio longer than the per-clip cap of the selected model and Seedance is selected (seedance2, seedance2-mini, or seedance2-5), do NOT clamp to 15s and drop the rest — split the run into multiple sound_to_video calls in the same turn (one per 15s segment, so a 20s audio becomes two clips: audioStart=0 duration=15, then audioStart=15 duration=5) and finish with a single stitch_video call referencing the resulting clip indices in order with audioIndex pointing at the same uploaded audio so the stitched output carries the full original soundtrack. LTX/WAN models accept up to 20s per clip, so single-call is fine for them. Use videoModel=\"wan3.0-video\" when the user explicitly requests Wan 3 audio-driven video. Use videoModel=\"wan3.0-spicy-video\" for Wan 3.0 Enhanced audio-driven video through MuleRouter provider ID w3.0-video; it supports adaptive ratios and provider prompt expansion.",
         "parameters": {
           "type": "object",
           "properties": {
@@ -1227,10 +1256,6 @@
               "minimum": 2,
               "maximum": 30
             },
-            "smartDuration": {
-              "type": "boolean",
-              "description": "Wan 3 only. Let the model choose 2-30 seconds. Do not also set duration. Sogni reserves 30 seconds and settles down to the provider-reported duration."
-            },
             "ratio": {
               "type": "string",
               "enum": [
@@ -1269,7 +1294,7 @@
                 "wan3.0-video",
                 "wan3.0-spicy-video"
               ],
-              "description": "\"ltx25-ia2v\" (default with image) and \"ltx25-a2v\" (default without image): LTX 2.5 image+audio and audio-only modes; Fast, HQ, and Pro currently use the release-validated official Distilled INT8 workflows. The Dev checkpoints are not publicly routed until upstream publishes and Sogni validates official ComfyUI Dev recipes. Video model. \"ltx23-ia2v\" (rollback with image): LTX 2.3 image+audio to video, audio-reactive with a reference image; Fast/HQ use the distilled 8-step worker and Default Media Quality Pro uses the non-distilled dev worker. \"ltx23-a2v\" (rollback without image): LTX 2.3 audio-only to video, no image needed, creates video purely from text prompt + audio with the same quality-tier routing. \"wan-s2v\": WAN 2.2 sound-to-video, best for lip-sync with a face image, fast 4-step. \"seedance2\": full Seedance 2.0 audio-reference video, 4-15s. \"seedance2-mini\": Seedance 2.0 Mini, 720p cap, fastest/lower-cost Seedance option. Seedance quality is selected only by this model value: pick \"seedance2-mini\" for faster/lower-cost drafts or explicit Mini requests, and pick \"seedance2\" for full-quality Seedance or 1080p/4K. Do not infer the Seedance model from Default Media Quality Fast/HQ/Pro. \"seedance2-5\": Seedance 2.5, the newest Seedance generation — 480p and 720p ONLY (it cannot render 1080p or 4K), 4-30s per clip at a fixed 24 fps, native audio, first-and-last-frame conditioning, and a much larger reference budget than the 2.0 family: up to 30 images, 10 videos, and 10 audios, with up to 50 reference media files total, subject to those per-modality caps. Choose \"seedance2-5\" when the user asks for Seedance 2.5, wants a single continuous Seedance clip longer than 15s (2.5 renders up to 30s in one call instead of being split and stitched), or wants a first-and-last-frame Seedance transition. Keep \"seedance2\" for 1080p/4K requests, which Seedance 2.5 cannot satisfy. For Seedance audio-reference prompts, preserve exact spoken dialogue when the user supplied it, and assign @Image1/@Audio1 roles. If the user asks for speech without words, describe the vocal performance without inventing quoted dialogue. Treat lip-sync, voice cloning, and real-human reference behavior as provider-sensitive rather than guaranteed. Omit to auto-select based on whether an image is present. \"wan3.0-video\" is Alibaba Wan 3 and \"wan3.0-spicy-video\" is MuleRouter w3.0-video. Both render 2-30s or smart duration at fixed 30 fps with optional native audio, provider prompt expansion, 480p/720p/1080p, adaptive/fixed ratios, first/last frames, and up to 10 image/5 video/5 audio references. Only Alibaba wan3.0-video accepts document/web context and watermark. Frame anchors and loose references are mutually exclusive. Do not send negativePrompt; video references are loose conditioning for a new result, not source-video editing or extension."
+              "description": "\"ltx25-ia2v\" (default with image) and \"ltx25-a2v\" (default without image): LTX 2.5 image+audio and audio-only modes; Fast, HQ, and Pro currently use the release-validated official Distilled INT8 workflows. The Dev checkpoints are not publicly routed until upstream publishes and Sogni validates official ComfyUI Dev recipes. Video model. \"ltx23-ia2v\" (rollback with image): LTX 2.3 image+audio to video, audio-reactive with a reference image; Fast/HQ use the distilled 8-step worker and Default Media Quality Pro uses the non-distilled dev worker. \"ltx23-a2v\" (rollback without image): LTX 2.3 audio-only to video, no image needed, creates video purely from text prompt + audio with the same quality-tier routing. \"wan-s2v\": WAN 2.2 sound-to-video, best for lip-sync with a face image, fast 4-step. \"seedance2\": full Seedance 2.0 audio-reference video, 4-15s. \"seedance2-mini\": Seedance 2.0 Mini, 720p cap, fastest/lower-cost Seedance option. Seedance quality is selected only by this model value: pick \"seedance2-mini\" for faster/lower-cost drafts or explicit Mini requests, and pick \"seedance2\" for full-quality Seedance or 1080p/4K. Do not infer the Seedance model from Default Media Quality Fast/HQ/Pro. \"seedance2-5\": Seedance 2.5, the newest Seedance generation — 480p and 720p ONLY (it cannot render 1080p or 4K), 4-30s per clip at a fixed 24 fps, native audio, first-and-last-frame conditioning, and a much larger reference budget than the 2.0 family: up to 30 images, 10 videos, and 10 audios, with up to 50 reference media files total, subject to those per-modality caps. Choose \"seedance2-5\" when the user asks for Seedance 2.5, wants a single continuous Seedance clip longer than 15s (2.5 renders up to 30s in one call instead of being split and stitched), or wants a first-and-last-frame Seedance transition. Keep \"seedance2\" for 1080p/4K requests, which Seedance 2.5 cannot satisfy. For Seedance audio-reference prompts, preserve exact spoken dialogue when the user supplied it, and assign @Image1/@Audio1 roles. If the user asks for speech without words, describe the vocal performance without inventing quoted dialogue. Treat lip-sync, voice cloning, and real-human reference behavior as provider-sensitive rather than guaranteed. Omit to auto-select based on whether an image is present. \"wan3.0-video\" is Alibaba Wan 3 and \"wan3.0-spicy-video\" is MuleRouter w3.0-video. Both render 2-30s at fixed 30 fps with optional native audio, provider prompt expansion, 480p/720p/1080p, adaptive/fixed ratios, first/last frames, and up to 10 image/5 video/5 audio references. Only Alibaba wan3.0-video accepts document/web context and watermark. Frame anchors and loose references are mutually exclusive. Do not send negativePrompt; video references are loose conditioning for a new result, not source-video editing or extension."
             },
             "generateAudio": {
               "type": "boolean",

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@sogni-ai/sogni-protocol",
-  "version": "1.0.0-alpha.36",
+  "version": "1.0.0-alpha.39",
   "description": "Language-neutral protocol artifacts for the Sogni ecosystem: tool schemas, prompts, OpenAI tool manifests, and enums. Consumed by every Sogni SDK (TypeScript, Swift, and future Python/Kotlin/Rust SDKs) so contracts stay in lockstep across languages.",
   "keywords": [
     "sogni",

package/prompts/tools/animate_photo.json

@@ -1,13 +1,14 @@
 {
   "contractId": "animate_photo_v1",
-  "version": "1.0.0",
+  "version": "1.1.0",
   "toolName": "animate_photo",
-  "baseDescription": "animate_photo produces video from one or more source images using LTX 2.5 by default, with LTX 2.3 retained as a rollback path.\n\nVIDEO PROMPT QUOTING: In video prompts, ONLY use double quotes for spoken dialogue.\nSpeaker tags are allowed outside the quotes for screenplay-style dialogue, e.g.\nCHARACTER: \"We made it.\" Never put on-screen text, overlay text, titles, captions, signs,\nwatermarks, or any visual text in quotes — describe them without quotes (e.g. bold white text\nreading CONGRATULATIONS overlays the lower third). Quotes signal speech to the model;\nquoting non-speech text confuses audio generation.\n\nDIALOGUE DURATION: Spoken dialogue in video prompts must fit the clip duration. Estimate\nat 2.5 words per second for natural cinematic delivery, plus ~1 second per acting beat\n(pauses, gestures, glances between lines). If the user did NOT explicitly request a specific\nduration (using default 5s), extend the duration to fit the dialogue (max 20s). If the user\nexplicitly requested a specific duration, condense the dialogue to fit while preserving meaning.\nAlways check: total dialogue words ÷ 2.5 + beat count ≤ clip duration.\n\nLATEST GENERATED IMAGE FOLLOW-UP: When the newest user turn asks to animate, make a video,\nor make a clip from a generated image/result (for example \"the apple\", \"this one\",\n\"the latest image\"), use animate_photo with that latest generated image. Do not inherit an\nolder Seedance model, resolution, or duration from an unrelated prior turn unless the newest\nuser turn explicitly says Seedance or confirms an immediately suggested Seedance video stage.\nLTX supports exact 2-20s durations, so honor requests like 3s exactly.\n\nWORD BUDGET PER CLIP: The handler REJECTS clips whose spoken dialogue exceeds the budget\n— there is NO auto-trim, so plan dialogue lengths up-front. Hard maximum is 3.75 spoken\nwords per second. Ceilings: 5s = 18 words, 6s = 22 words, 8s = 30 words, 10s = 37 words,\n15s = 56 words, 20s = 75 words. Aim below these ceilings. If a scene's dialogue won't fit,\ntighten the lines, raise the per-clip duration, or split into two segments — do NOT submit\nand hope it works. Spoken words inside double quotes count toward the budget; speaker tags\nand visual/action prose are free.\n\nBATCH VIDEO PER-CLIP DURATION: For a multi-segment animate_photo batch\n(sourceImageIndices + prompts) when the user states a TOTAL video length but NO per-clip\nlength, target 15 seconds per clip when dialogue is involved, and pass that duration\nexplicitly. Example: 60s total → 4 segments × 15s, NOT 6×10s or 12×5s. There is NO 3-clip\nbatch cap: sourceImageIndices supports up to 16 clips, so never split one planned batch into\n\"first 3\" and \"remaining clips\" calls. Do NOT split a planned 15s dialogue scene into multiple\nshorter clips just because a retry complains about word budget; keep duration=15 and tighten\nthe line. Use 5s clips only for single short motion beats or one very short spoken phrase.\nIf the user explicitly specifies a per-clip duration, honor that instead.\n\nN-VERSIONS-OF-A-VIDEO PATTERN: NEVER call animate_photo N times sequentially — ALWAYS\nuse sourceImageIndices in ONE call so all N projects run in parallel. Two flavors:\n(A) SHARED CONTENT — one edit_image/generate_image call with numberOfVariations=N + {|}\nDynamic Prompts to make N distinct source images, then ONE animate_photo call with\nsourceImageIndices=[start..start+N-1] and a single shared prompt.\n(B) PER-CLIP CONTENT — when each clip has DIFFERENT dialogue, jokes, narration, or motion,\npass BOTH sourceImageIndices AND prompts (array of N strings, one per clip) in the SAME\nsingle animate_photo call. The top-level prompt is still required — pass a brief batch summary.\n\nCRITICAL: sourceImageIndices values MUST be read from the latest edit_image/generate_image\ntool result's startIndex field — if startIndex=3 and 4 images were generated, pass\nsourceImageIndices=[3,4,5,6], NOT [0,1,2,3]. Negative indices refer to uploaded images:\n-1 first upload, -2 second upload, -3 third upload. Use repeated -1 entries only when\nintentionally reusing the primary uploaded image. When prompts is supplied, prompts.length\nMUST equal sourceImageIndices.length.\n\nSEEDANCE UPLOADED STORYBOARD DEFAULT: If the user uploaded a storyboard, shot sheet,\nor visual trailer board and asks to make a trailer/video/movie/clip from it, do NOT use\nanimate_photo on the board image and do NOT split it into four LTX clips. Use generate_video\nwith Seedance referenceImageIndices for one continuous clip unless the user explicitly asks\nfor separate LTX clips or first-frame/last-frame animation.\n\nSCREENPLAY / STORYBOARD ANIMATE RULE: For full storyboard projects, use one\nanimate_photo batch with sourceImageIndices + prompts so each clip keeps its own exact\nscene text, stable cast anchors, and screenplay-style speaker-tagged dialogue, and all video\nclips render in parallel. Every speaking clip's video prompt must include that clip's actual\nquoted dialogue, not placeholders such as \"while speaking\", \"dialogue begins\", \"explaining\",\nor \"final line lands\". If each generated scene keyframe should be both the first and last frame\nof its own stitched segment, call animate_photo with sourceImageIndices=[start..end],\nframeRole=\"both\", prompts=[...], and OMIT endImageIndex/endImageIndices so the handler\nuses each source as its own end frame.\n\nUPLOADED REFERENCE LOOPED SKITS: When the user supplies one uploaded reference image and\nasks for several scripted/storyboard/dialogue segments to reuse that same image as BOTH the\nfirst frame and last frame of each segment before stitching, do it in ONE animate_photo call:\nsourceImageIndices=[-1,-1,...], frameRole=\"both\", endImageIndex=-1 (or matching\nendImageIndices=[-1,-1,...]), duration equal to the requested per-segment duration, and\nprompts=[one full scene prompt per segment]. Each prompt must preserve the exact screenplay\nspeaker tags and quoted dialogue from that scene, e.g. HOST: \"...\" GUEST: \"...\". Do not\ndrop speaker tags, convert them to generic narration, omit the last-frame contract, analyze\nthe image first, generate new keyframes first, or split the batch into serial calls. After\nthe single animate_photo batch completes, call stitch_video with the returned video indices.\n\nFor adjacent transition chains: N images create N-1 clips — call animate_photo with\nframeRole=\"both\", sourceImageIndices=[start..end-1], endImageIndices=[start+1..end],\nprompts=[one transition prompt per adjacent pair], then stitch_video. If 5 uploaded images\nare the keyframe sequence, use sourceImageIndices=[-1,-2,-3,-4],\nendImageIndices=[-2,-3,-4,-5], frameRole=\"both\", prompts length 4, then stitch_video.\nDo NOT set endImageIndex=-1 in generated-keyframe patterns — that means every clip ends\non the primary uploaded image.\n\nUPLOADED FIRST-FRAME/LAST-FRAME TRANSITION CHAINS: If the user uploads multiple images\nand asks for a video that transitions from image to image, changes country/version every\nN seconds, or says to use first-frame/last-frame for each pair, call animate_photo directly.\nDo not call edit_image, generate_image, analyze_image, or map_assets_for_model first — the\nuploaded images are already the keyframes. For N uploaded images, create N-1 adjacent clips\nunless the user explicitly asks for a loop back to the first image. Use per-clip duration\nfrom \"every N seconds\" when present; otherwise divide the requested total by the number of\nadjacent clips. After animate_photo returns the batch videos, always call stitch_video with\nthose video indices before finalizing.",
+  "baseDescription": "animate_photo produces video from one or more source images using LTX 2.5 by default, with LTX 2.3 retained as a rollback path.\n\nVIDEO PROMPT QUOTING: In video prompts, ONLY use double quotes for spoken dialogue.\nSpeaker tags are allowed outside the quotes for screenplay-style dialogue, e.g.\nCHARACTER: \"We made it.\" Never put on-screen text, overlay text, titles, captions, signs,\nwatermarks, or any visual text in quotes — describe them without quotes (e.g. bold white text\nreading CONGRATULATIONS overlays the lower third). Quotes signal speech to the model;\nquoting non-speech text confuses audio generation.\n\nDIALOGUE DURATION: Spoken dialogue in video prompts must fit the clip duration. Estimate\nat 2.5 words per second for natural cinematic delivery, plus ~1 second per acting beat\n(pauses, gestures, glances between lines). If the user did NOT explicitly request a specific\nduration (using default 5s), extend the duration to fit the dialogue (max 20s). If the user\nexplicitly requested a specific duration, condense the dialogue to fit while preserving meaning.\nAlways check: total dialogue words ÷ 2.5 + beat count ≤ clip duration.\n\nLATEST GENERATED IMAGE FOLLOW-UP: When the newest user turn asks to animate, make a video,\nor make a clip from a generated image/result (for example \"the apple\", \"this one\",\n\"the latest image\"), use animate_photo with that latest generated image. Do not inherit an\nolder Seedance model, resolution, or duration from an unrelated prior turn unless the newest\nuser turn explicitly says Seedance or confirms an immediately suggested Seedance video stage.\nLTX supports exact 2-20s durations, so honor requests like 3s exactly.\n\nWORD BUDGET PER CLIP: The handler REJECTS clips whose spoken dialogue exceeds the budget\n— there is NO auto-trim, so plan dialogue lengths up-front. Hard maximum is 3.75 spoken\nwords per second. Ceilings: 5s = 18 words, 6s = 22 words, 8s = 30 words, 10s = 37 words,\n15s = 56 words, 20s = 75 words. Aim below these ceilings. If a scene's dialogue won't fit,\ntighten the lines, raise the per-clip duration, or split into two segments — do NOT submit\nand hope it works. Spoken words inside double quotes count toward the budget; speaker tags\nand visual/action prose are free.\n\nBATCH VIDEO PER-CLIP DURATION: For a multi-segment animate_photo batch\n(sourceImageIndices + prompts) when the user states a TOTAL video length but NO per-clip\nlength, target 15 seconds per clip when dialogue is involved, and pass that duration\nexplicitly. Example: 60s total → 4 segments × 15s, NOT 6×10s or 12×5s. There is NO 3-clip\nbatch cap: sourceImageIndices supports up to 16 clips, so never split one planned batch into\n\"first 3\" and \"remaining clips\" calls. Do NOT split a planned 15s dialogue scene into multiple\nshorter clips just because a retry complains about word budget; keep duration=15 and tighten\nthe line. Use 5s clips only for single short motion beats or one very short spoken phrase.\nIf the user explicitly specifies a per-clip duration, honor that instead.\n\nN-VERSIONS-OF-A-VIDEO PATTERN: NEVER call animate_photo N times sequentially — ALWAYS\nuse sourceImageIndices in ONE call so all N projects run in parallel. Two flavors:\n(A) SHARED CONTENT — one edit_image/generate_image call with numberOfVariations=N + {|}\nDynamic Prompts to make N distinct source images, then ONE animate_photo call with\nsourceImageIndices=[start..start+N-1] and a single shared prompt.\n(B) PER-CLIP CONTENT — when each clip has DIFFERENT dialogue, jokes, narration, or motion,\npass BOTH sourceImageIndices AND prompts (array of N strings, one per clip) in the SAME\nsingle animate_photo call. The top-level prompt is still required — pass a brief batch summary.\n\nCRITICAL: sourceImageIndices values MUST be read from the latest edit_image/generate_image\ntool result's startIndex field — if startIndex=3 and 4 images were generated, pass\nsourceImageIndices=[3,4,5,6], NOT [0,1,2,3]. Negative indices refer to uploaded images:\n-1 first upload, -2 second upload, -3 third upload. Use repeated -1 entries only when\nintentionally reusing the primary uploaded image. When prompts is supplied, prompts.length\nMUST equal sourceImageIndices.length.\n\nSEEDANCE UPLOADED STORYBOARD DEFAULT: If the user uploaded a storyboard, shot sheet,\nor visual trailer board and asks to make a trailer/video/movie/clip from it, do NOT use\nanimate_photo on the board image and do NOT split it into four LTX clips. Use generate_video\nwith Seedance referenceImageIndices for one continuous clip unless the user explicitly asks\nfor separate LTX clips or first-frame/last-frame animation.\n\nSCREENPLAY / STORYBOARD ANIMATE RULE: For full storyboard projects, use one\nanimate_photo batch with sourceImageIndices + prompts so each clip keeps its own exact\nscene text, stable cast anchors, and screenplay-style speaker-tagged dialogue, and all video\nclips render in parallel. Every speaking clip's video prompt must include that clip's actual\nquoted dialogue, not placeholders such as \"while speaking\", \"dialogue begins\", \"explaining\",\nor \"final line lands\". If each generated scene keyframe should be both the first and last frame\nof its own stitched segment, call animate_photo with sourceImageIndices=[start..end],\nframeRole=\"both\", prompts=[...], and OMIT endImageIndex/endImageIndices so the handler\nuses each source as its own end frame.\n\nUPLOADED REFERENCE LOOPED SKITS: When the user supplies one uploaded reference image and\nasks for several scripted/storyboard/dialogue segments to reuse that same image as BOTH the\nfirst frame and last frame of each segment before stitching, do it in ONE animate_photo call:\nsourceImageIndices=[-1,-1,...], frameRole=\"both\", endImageIndex=-1 (or matching\nendImageIndices=[-1,-1,...]), duration equal to the requested per-segment duration, and\nprompts=[one full scene prompt per segment]. Each prompt must preserve the exact screenplay\nspeaker tags and quoted dialogue from that scene, e.g. HOST: \"...\" GUEST: \"...\". Do not\ndrop speaker tags, convert them to generic narration, omit the last-frame contract, analyze\nthe image first, generate new keyframes first, or split the batch into serial calls. After\nthe single animate_photo batch completes, call stitch_video with the returned video indices.\n\nFor adjacent transition chains: N images create N-1 clips — call animate_photo with\nframeRole=\"both\", sourceImageIndices=[start..end-1], endImageIndices=[start+1..end],\nprompts=[one transition prompt per adjacent pair], then stitch_video. If 5 uploaded images\nare the keyframe sequence, use sourceImageIndices=[-1,-2,-3,-4],\nendImageIndices=[-2,-3,-4,-5], frameRole=\"both\", prompts length 4, then stitch_video.\nDo NOT set endImageIndex=-1 in generated-keyframe patterns — that means every clip ends\non the primary uploaded image.\n\nUPLOADED FIRST-FRAME/LAST-FRAME TRANSITION CHAINS: If the user uploads multiple images\nand asks for a video that transitions from image to image, changes country/version every\nN seconds, or says to use first-frame/last-frame for each pair, call animate_photo directly.\nDo not call edit_image, generate_image, analyze_image, or map_assets_for_model first — the\nuploaded images are already the keyframes. For N uploaded images, create N-1 adjacent clips\nunless the user explicitly asks for a loop back to the first image. Use per-clip duration\nfrom \"every N seconds\" when present; otherwise divide the requested total by the number of\nadjacent clips. After animate_photo returns the batch videos, always call stitch_video with\nthose video indices before finalizing.\n\nMINIMAX H3 2K OUTPUT: outputScale=2 is available only on the MiniMax H3 selectors. It renders on the normal H3 canvas and delivers the clip at twice its width and height (1344x768 becomes 2688x1536) with the same length and audio, for +10 Spark per second (+6 at 480p). Set outputScale=2 only when the user asks for 2K, 1440p-class or extra-sharp H3 output; keep targetResolution at 768 or omit it, never set 1080p, 1440p or 4K for H3, and leave outputScale unset otherwise. Do not set it for any other model.",
   "parameterDocs": {
     "sourceImageIndices": "Batch source image indices. Read startIndex from prior generate_image/edit_image result. Negative = uploaded images (-1 = first upload).",
     "prompts": "Per-clip prompt array. Length MUST equal sourceImageIndices.length when both are set.",
     "duration": "Per-clip duration in seconds. Target 15s when dialogue is involved and total length is given without per-clip spec.",
     "frameRole": "Set to \"both\" for first+last frame transitions using sourceImageIndices + endImageIndices.",
-    "endImageIndices": "End frames for adjacent-chain transitions. N images → N-1 clips."
+    "endImageIndices": "End frames for adjacent-chain transitions. N images → N-1 clips.",
+    "outputScale": "MiniMax H3 only: 2 = 2K delivery (twice the width and height, same length and audio, +10 Spark/s, +6 at 480p). Set only when the user asks for 2K, 1440p-class or extra-sharp H3 output; otherwise omit."
   }
 }

package/prompts/tools/generate_video.json

@@ -1,10 +1,11 @@
 {
   "contractId": "generate_video_v1",
-  "version": "1.1.0",
+  "version": "1.2.0",
   "toolName": "generate_video",
-  "baseDescription": "generate_video produces text-to-video clips and Seedance multimodal reference videos.\nUse for text-only video generation with no source image input. For Seedance, also use this\ntool when uploaded/generated images, videos, or audio are loose references. Use animate_photo\nonly when a non-Seedance source image must become the first frame of an LTX/WAN animation.\n\nSEEDANCE UPLOADED STORYBOARD DEFAULT: When the user uploads a storyboard, shot sheet,\nmood board, or trailer concept image and asks to make a movie trailer/video/clip from it,\ndefault to one Seedance generate_video call with referenceImageIndices=[-1]. Do not first\nextract panels with edit_image, do not generate replacement keyframes, and do not make four\nseparate LTX animate_photo clips unless the user explicitly asks for separate clips or LTX.\nUse seedance2 when premium Spark access is available; if premium access is unavailable,\nexplain the limitation or use the best non-Seedance fallback the user accepts.\n\nSTORYTELLING / COMMERCIAL / TRAILER PROMPTS: For creative video requests, turn the brief\ninto timed, causally connected visual beats before writing the final prompt. Default social\nvideo is 15s 9:16 with a strong first 1-2s, visible escalation, payoff, and brand/CTA/final\nimage. Commercials should show audience desire/problem, transformation, proof/benefit, and\nCTA. Trailers should follow hook → world → disruption → escalation → reveal → title/CTA.\nEvery beat must be generatable: subject, setting, action, camera, lighting, audio, and text\nrole where relevant. Avoid vague \"cinematic\" filler, feature dumps, and beautiful images with\nno visible change.\n\nVIDEO PROMPT QUOTING: ONLY use double quotes for spoken dialogue in video prompts. Never\nquote on-screen text, titles, captions, or visual text elements — describe them without\nquotes. Quotes signal speech to the model and confuse audio generation.\n\nSTORYBOARD TEXT: Structural headings, section numbers, slide titles, panel titles, and\ncaptions in storyboard references may become short audio-only narration/VO or\nkey-message beats, but they are not subtitles, title cards, lower thirds, or visible\noverlays unless the user explicitly asks for visible text, on-screen text, a title\ncard, subtitle, lower third, signage, or CTA. Keep narration as separate brief phrases\nwith pauses; do not concatenate storyboard labels into run-on voiceover.\n\nDIALOGUE DURATION: Spoken dialogue must fit the clip. Estimate 2.5 words per second\nnatural delivery plus ~1s per acting beat. Hard maximum 3.75 words/second.\nCheck: dialogue words ÷ 2.5 + beats ≤ duration. Do not submit oversized dialogue.\n\nLATEST USER DURATION WINS: In follow-up turns, use the newest duration the user states,\neven if a previous assistant message mentioned a longer script/runtime. For example, if\nhistory says \"the full script is 66 seconds\" but the user now says \"do a 30 second version\",\ngenerate the 30 second version. Do not ask a clarification question just because history\ncontains another duration; treat the latest user request as the override.\n\nSEEDANCE DURATION LIMITS: Seedance 2.0 and Mini support 4-15s clips; Seedance 2.5 supports 4-30s clips. If the user explicitly asks\nfor Seedance below 4s, do not silently round up. Ask whether they prefer a 4s Seedance clip\nor an exact-duration LTX clip. If the user did not explicitly ask for Seedance, choose the\nmodel/tool that can satisfy the requested duration exactly.",
+  "baseDescription": "generate_video produces text-to-video clips and Seedance multimodal reference videos.\nUse for text-only video generation with no source image input. For Seedance, also use this\ntool when uploaded/generated images, videos, or audio are loose references. Use animate_photo\nonly when a non-Seedance source image must become the first frame of an LTX/WAN animation.\n\nSEEDANCE UPLOADED STORYBOARD DEFAULT: When the user uploads a storyboard, shot sheet,\nmood board, or trailer concept image and asks to make a movie trailer/video/clip from it,\ndefault to one Seedance generate_video call with referenceImageIndices=[-1]. Do not first\nextract panels with edit_image, do not generate replacement keyframes, and do not make four\nseparate LTX animate_photo clips unless the user explicitly asks for separate clips or LTX.\nUse seedance2 when premium Spark access is available; if premium access is unavailable,\nexplain the limitation or use the best non-Seedance fallback the user accepts.\n\nSTORYTELLING / COMMERCIAL / TRAILER PROMPTS: For creative video requests, turn the brief\ninto timed, causally connected visual beats before writing the final prompt. Default social\nvideo is 15s 9:16 with a strong first 1-2s, visible escalation, payoff, and brand/CTA/final\nimage. Commercials should show audience desire/problem, transformation, proof/benefit, and\nCTA. Trailers should follow hook → world → disruption → escalation → reveal → title/CTA.\nEvery beat must be generatable: subject, setting, action, camera, lighting, audio, and text\nrole where relevant. Avoid vague \"cinematic\" filler, feature dumps, and beautiful images with\nno visible change.\n\nVIDEO PROMPT QUOTING: ONLY use double quotes for spoken dialogue in video prompts. Never\nquote on-screen text, titles, captions, or visual text elements — describe them without\nquotes. Quotes signal speech to the model and confuse audio generation.\n\nSTORYBOARD TEXT: Structural headings, section numbers, slide titles, panel titles, and\ncaptions in storyboard references may become short audio-only narration/VO or\nkey-message beats, but they are not subtitles, title cards, lower thirds, or visible\noverlays unless the user explicitly asks for visible text, on-screen text, a title\ncard, subtitle, lower third, signage, or CTA. Keep narration as separate brief phrases\nwith pauses; do not concatenate storyboard labels into run-on voiceover.\n\nDIALOGUE DURATION: Spoken dialogue must fit the clip. Estimate 2.5 words per second\nnatural delivery plus ~1s per acting beat. Hard maximum 3.75 words/second.\nCheck: dialogue words ÷ 2.5 + beats ≤ duration. Do not submit oversized dialogue.\n\nLATEST USER DURATION WINS: In follow-up turns, use the newest duration the user states,\neven if a previous assistant message mentioned a longer script/runtime. For example, if\nhistory says \"the full script is 66 seconds\" but the user now says \"do a 30 second version\",\ngenerate the 30 second version. Do not ask a clarification question just because history\ncontains another duration; treat the latest user request as the override.\n\nSEEDANCE DURATION LIMITS: Seedance 2.0 and Mini support 4-15s clips; Seedance 2.5 supports 4-30s clips. If the user explicitly asks\nfor Seedance below 4s, do not silently round up. Ask whether they prefer a 4s Seedance clip\nor an exact-duration LTX clip. If the user did not explicitly ask for Seedance, choose the\nmodel/tool that can satisfy the requested duration exactly.\n\nMINIMAX H3 2K OUTPUT: outputScale=2 is available only on the MiniMax H3 selectors. It renders on the normal H3 canvas and delivers the clip at twice its width and height (1344x768 becomes 2688x1536) with the same length and audio, for +10 Spark per second (+6 at 480p). Set outputScale=2 only when the user asks for 2K, 1440p-class or extra-sharp H3 output; keep targetResolution at 768 or omit it, never set 1080p, 1440p or 4K for H3, and leave outputScale unset otherwise. Do not set it for any other model.",
   "parameterDocs": {
     "prompt": "Video prompt. Use double quotes ONLY for spoken dialogue. Describe visual text without quotes.",
-    "duration": "Clip duration in seconds. Plan dialogue word count against the 3.75 words/second ceiling."
+    "duration": "Clip duration in seconds. Plan dialogue word count against the 3.75 words/second ceiling.",
+    "outputScale": "MiniMax H3 only: 2 = 2K delivery (twice the width and height, same length and audio, +10 Spark/s, +6 at 480p). Set only when the user asks for 2K, 1440p-class or extra-sharp H3 output; otherwise omit."
   }
 }

package/prompts/tools/upscale_video.json

@@ -1,10 +1,13 @@
 {
   "contractId": "upscale_video_v1",
-  "version": "1.0.0",
+  "version": "1.1.0",
   "toolName": "upscale_video",
-  "baseDescription": "upscale_video performs promptless, deterministic FlashVSR super-resolution of exactly one\nuploaded or generated video. Use it when the user asks to upscale, enlarge, sharpen, or increase\nthe resolution of an existing video, or wants a 1080p, 1440p, 2K, or HD copy of it. The output\nkeeps every source frame, the exact frame rate, the full aspect ratio, and the original audio.\n\nDo not invent a prompt and do not route a pure video upscale through video_to_video,\ngenerate_video, extend_video, or replace_video_segment; those tools re-render or edit content.\nUse video_to_video only when the user explicitly names a generative model such as Seedance for\nthe re-render. Do not use upscale_image for videos.\n\nOmit sourceVideoIndex to use the latest generated video, falling back to the most recent upload.\nUse zero-based non-negative indices for generated videos; -1 selects the first uploaded video\nand -2 the second. targetResolution is the output short edge: omit it for 1440p (or 1080p when\nthe source is too small for 1440p), and set 1080 when the user asks for 1080p or Full HD.\nThe output is at most twice the source size, so 1440p needs a source short edge of 720-768px\nand 1080p needs 540-768px. Sources must also be at most about 1344x768 pixels (768x1344 in\nportrait), 1-60 fps, and 100 MB. The server sets the maximum clip length and rejects a source\nthat is too long; never quote a length limit yourself. If the tool reports that the source is too\nlong or outside these limits, relay that error to the user instead of switching to a generative\nvideo tool. It cannot produce 4K or any size above 1440p: when the user asks for one, say so and\noffer 1440p instead of silently delivering less.",
+  "baseDescription": "upscale_video performs promptless, deterministic FlashVSR super-resolution of exactly one\nuploaded or generated video. Use it when the user asks to upscale, enlarge, sharpen, or increase\nthe resolution of an existing video, or wants a 1080p, 1440p, 2K, or HD copy of it. The output\nkeeps every source frame, the exact frame rate, the full aspect ratio, and the original audio.\n\nDo not invent a prompt and do not route a pure video upscale through video_to_video,\ngenerate_video, extend_video, or replace_video_segment; those tools re-render or edit content.\nUse video_to_video only when the user explicitly names a generative model such as Seedance for\nthe re-render. Do not use upscale_image for videos.\n\nOmit sourceVideoIndex to use the latest generated video, falling back to the most recent upload.\nUse zero-based non-negative indices for generated videos; -1 selects the first uploaded video\nand -2 the second. targetResolution is the output short edge: omit it for 1440p (or 1080p when\nthe source is too small for 1440p), and set 1080 when the user asks for 1080p or Full HD.\nThe output is at most twice the source size, so 1440p needs a source short edge of 720-768px\nand 1080p needs 540-768px. Sources must also be at most about 1344x768 pixels (768x1344 in\nportrait), 1-60 fps, and 100 MB. The server sets the maximum clip length and rejects a source\nthat is too long; never quote a length limit yourself. If the tool reports that the source is too\nlong or outside these limits, relay that error to the user instead of switching to a generative\nvideo tool. It cannot produce 4K or any size above 1440p: when the user asks for one, say so and\noffer 1440p instead of silently delivering less.\n\nThe defaults give the most stable result. Set detailPreference to sharper only when the user\nasks for a sharper, crisper, or more detailed upscale, and processingSpeed to faster only when\nthey ask for a quicker one. Leave seed out unless the user gives a seed or asks for a different\nor random variation (-1 is random). None of these options changes the price.",
   "parameterDocs": {
     "sourceVideoIndex": "Omit for the latest generated video, then the most recent upload. Generated videos are zero-based; -1/-2 select uploaded videos.",
-    "targetResolution": "Output short edge in pixels: 1440 or 1080. Omit for the default (1440, or 1080 for sources under 720px); set 1080 when the user asks for 1080p or Full HD."
+    "targetResolution": "Output short edge in pixels: 1440 or 1080. Omit for the default (1440, or 1080 for sources under 720px); set 1080 when the user asks for 1080p or Full HD.",
+    "detailPreference": "stable (default) or sharper. Set sharper only when the user asks for a sharper, crisper, or more detailed result.",
+    "processingSpeed": "stable (default) or faster. Set faster only when the user asks for a quicker upscale.",
+    "seed": "Omit for the repeatable default 0. -1 is random; otherwise an integer 0-4294967295 the user gave."
   }
 }

package/schemas/tools/animate_photo.schema.json

@@ -36,7 +36,7 @@
         "wan3.0-video",
         "wan3.0-spicy-video"
       ],
-      "description": "\"ltx25\" (default): LTX 2.5 I2V or first/last-frame video with native audio; Fast, HQ, and Pro currently use the release-validated official Distilled INT8 workflow. The Dev checkpoints are not publicly routed until upstream publishes and Sogni validates an official ComfyUI Dev recipe. Which video model to use. \"ltx23\": LTX 2.3 rollback with native audio. \"wan22\": quick simple motion without audio, up to 10s. \"minimax-h3-i2v\" is standard MiniMax H3 from one first frame; \"minimax-h3-i2v-turbo\" is the existing 4-step LightX2V Turbo engine; \"minimax-h3-fasth3-i2v-turbo\" is the separate FastVideo VSA four-step FastH3 engine, about 2x faster and fixed to Euler/simple. The matching FLF2V selectors provide standard, LightX2V Turbo, and FastH3 Turbo first/last-frame generation; FastH3 has no R2V mode; use frameRole=\"both\" and provide the end frame. H3 generates native audio at fixed 24fps for 5.17-15.08s and has no negative-prompt input. H3 Base and Turbo prompts use the exact three-field contract and the official mode-specific alignment line. Do not set Seedance here; use generate_video with Seedance references. \"wan3.0-video\" is Alibaba Wan 3 and \"wan3.0-spicy-video\" is MuleRouter w3.0-video. Both render 2-30s or smart duration at fixed 30 fps with optional native audio, provider prompt expansion, 480p/720p/1080p, adaptive/fixed ratios, first/last frames, and up to 10 image/5 video/5 audio references. Only Alibaba wan3.0-video accepts document/web context and watermark. Frame anchors and loose references are mutually exclusive. Do not send negativePrompt; video references are loose conditioning for a new result, not source-video editing or extension."
+      "description": "\"ltx25\" (default): LTX 2.5 I2V or first/last-frame video with native audio; Fast, HQ, and Pro currently use the release-validated official Distilled INT8 workflow. The Dev checkpoints are not publicly routed until upstream publishes and Sogni validates an official ComfyUI Dev recipe. Which video model to use. \"ltx23\": LTX 2.3 rollback with native audio. \"wan22\": quick simple motion without audio, up to 10s. \"minimax-h3-i2v\" is standard MiniMax H3 from one first frame; \"minimax-h3-i2v-turbo\" is the existing 4-step LightX2V Turbo engine; \"minimax-h3-fasth3-i2v-turbo\" is the separate FastVideo VSA four-step FastH3 engine, about 2x faster and fixed to Euler/simple. The matching FLF2V selectors provide standard, LightX2V Turbo, and FastH3 Turbo first/last-frame generation; FastH3 has no R2V mode; use frameRole=\"both\" and provide the end frame. H3 generates native audio at fixed 24fps for 5.17-15.08s and has no negative-prompt input. H3 Base and Turbo prompts use the exact three-field contract and the official mode-specific alignment line. Do not set Seedance here; use generate_video with Seedance references. \"wan3.0-video\" is Alibaba Wan 3 and \"wan3.0-spicy-video\" is MuleRouter w3.0-video. Both render 2-30s at fixed 30 fps with optional native audio, provider prompt expansion, 480p/720p/1080p, adaptive/fixed ratios, first/last frames, and up to 10 image/5 video/5 audio references. Only Alibaba wan3.0-video accepts document/web context and watermark. Frame anchors and loose references are mutually exclusive. Do not send negativePrompt; video references are loose conditioning for a new result, not source-video editing or extension."
     },
     "negativePrompt": {
       "type": "string",
@@ -50,10 +50,6 @@
       "type": "number",
       "description": "Video duration in seconds. Default: 5. Use when the user explicitly requests a specific length (e.g., \"make a 10 second video\"). Per-model maximum: ltx25 and ltx23 = 20s, wan22 = 10s (clips longer than this are invalid), wan3.0-video = 30s with a 2s minimum, minimax-h3 = 15.08s with a 5.17s minimum because H3 renders 124-362 frames on a 17-frame grid at a fixed 24 fps. For totals beyond the per-model cap, batch multiple clips via sourceImageIndices instead of requesting a single oversized clip."
     },
-    "smartDuration": {
-      "type": "boolean",
-      "description": "Wan 3 only. Let the model choose 2-30 seconds. Do not also set duration. The 30-second maximum is reserved and the final charge settles down to reported duration."
-    },
     "ratio": {
       "type": "string",
       "enum": [
@@ -72,7 +68,15 @@
     },
     "targetResolution": {
       "type": "number",
-      "description": "Short-side video resolution target in pixels. Use when the user asks for a bare named resolution such as \"480p\", \"720p\", or \"1080p\" without exact pixels or an output orientation. This preserves the source image aspect ratio. Wan 3 supports 480p, 720p, and 1080p; HappyHorse supports only 720p and 1080p. Never set 4K for either. MiniMax H3 renders inside a 1344x768 pixel budget on a 32px grid, so use 768 for H3 and never 1080p or 4K. Do NOT set width, height, or exact-pixel aspectRatio for bare named resolution requests. If the user says \"720p portrait\" or \"720p landscape\", use exact-pixel aspectRatio instead."
+      "description": "Short-side video resolution target in pixels. Use when the user asks for a bare named resolution such as \"480p\", \"720p\", or \"1080p\" without exact pixels or an output orientation. This preserves the source image aspect ratio. Wan 3 supports 480p, 720p, and 1080p; HappyHorse supports only 720p and 1080p. Never set 4K for either. MiniMax H3 renders inside a 1344x768 pixel budget on a 32px grid, so use 768 for H3 and never 1080p or 4K. Do NOT set width, height, or exact-pixel aspectRatio for bare named resolution requests. If the user says \"720p portrait\" or \"720p landscape\", use exact-pixel aspectRatio instead. For MiniMax H3 2K requests keep targetResolution at 768 (or omit it) and set outputScale to 2 instead."
+    },
+    "outputScale": {
+      "type": "integer",
+      "enum": [
+        1,
+        2
+      ],
+      "description": "MiniMax H3 only. 2 delivers 2K output: the clip renders on the normal H3 canvas and is delivered at twice its width and height (1344x768 becomes 2688x1536) with the same length and audio, for +10 Spark per second (+6 at 480p). Set 2 only when the user asks for 2K, 1440p-class or extra-sharp H3 output; leave unset otherwise. Ignored for other models."
     },
     "sourceImageIndex": {
       "type": "number",
@@ -140,7 +144,7 @@
         "type": "string",
         "minLength": 1
       },
-      "description": "Ordered LoRA IDs to apply to a MiniMax H3 render. Use only when the user explicitly asks for a LoRA or for an effect one of these names describes. Stack up to 8 in one request; order matters because the adapters apply in sequence and do not commute. Keep this array positionally aligned with loraStrengths. The first render with an uncached LoRA takes longer to start while the worker downloads it.\n\nAccepted only when videoModel is one of \"minimax-h3-i2v\", \"minimax-h3-i2v-turbo\", \"minimax-h3-fasth3-i2v-turbo\", \"minimax-h3-flf2v\", \"minimax-h3-flf2v-turbo\", \"minimax-h3-fasth3-flf2v-turbo\". Every other video model on this tool loads no LoRAs and silently ignores these arrays, so set videoModel to an H3 mode in the same call when the user asks for one.\n\nOne LoRA is published for MiniMax H3 today: h3-realism-people (fal), a realism pass trained on live-action footage of people. It restores skin texture and pores, stray hairs, fabric weave and a fine sensor grain that the base model smooths away, and holds up in close-up. It needs its trigger word: put r34l1sm near the FRONT of the prompt, or the render comes back as ordinary H3 with no error. Exact ranges and any LoRA published since: GET /v1/loras/comfy?modelId=<model>. Do not invent ids."
+      "description": "Ordered LoRA IDs to apply to a MiniMax H3 render. Use only when the user explicitly asks for a LoRA or for an effect one of these names describes. Stack up to 8 in one request; order matters because the adapters apply in sequence and do not commute. Keep this array positionally aligned with loraStrengths. The first render with an uncached LoRA takes longer to start while the worker downloads it.\n\nAccepted only when videoModel is one of \"minimax-h3-i2v\", \"minimax-h3-i2v-turbo\", \"minimax-h3-fasth3-i2v-turbo\", \"minimax-h3-flf2v\", \"minimax-h3-flf2v-turbo\", \"minimax-h3-fasth3-flf2v-turbo\". Every other video model on this tool loads no LoRAs and silently ignores these arrays, so set videoModel to an H3 mode in the same call when the user asks for one.\n\nFive LoRAs are published for MiniMax H3 today and the set differs by mode, so GET /v1/loras/comfy?modelId=<model> is authoritative for the mode in hand and carries exact ranges, maturity flags, and anything published since. h3-realism-people (fal) is a realism pass trained on live-action footage of people: it restores skin texture and pores, stray hairs, fabric weave and a fine sensor grain that the base model smooths away, and holds up in close-up. It is the only one gated on a trigger word — put r34l1sm near the FRONT of the prompt, or the render comes back as ordinary H3 with no error. h3-vbvr-video-reasoning is a prompt-adherence pass that holds the model to what was asked instead of improvising. h3-natural-face-speech (AdaptiveVision) makes people talking on camera look and sound more natural: cheeks, brows, jaw and lips move together as in real speech, and spoken English comes through clearer; use it for talking-head shots such as vlogs, podcasts, interviews and presenters. h3-better-motion (AdaptiveVision) gives people more natural, consistent body movement — weight shifts, strides, turns and gestures that follow through — for dance, sport, walking and other full-body shots. Both AdaptiveVision LoRAs work best with short, simple prompt sentences and are not validated on reference-to-video. h3-mystic-xxx-v4 is an uncensored adult fine-tune. Do not invent ids."
     },
     "loraStrengths": {
       "type": "array",
@@ -149,7 +153,7 @@
       "items": {
         "type": "number"
       },
-      "description": "Strength for each LoRA in loras, in the same order. Omitting the array applies 1.0 to every LoRA, which is NOT the catalog default and for h3-realism-people is already at the top of its band, so send explicit values. Video LoRAs are positive-only — unlike the bipolar Krea 2 image sliders, a negative value is not an inverse effect and 0 is off. h3-realism-people takes 0-2 and its catalog default is 0.8; 0.6-1 is the usable band. It also pulls the camera in as it climbs: at 1.5 and above the shot reliably recomposes and the grade darkens, which on an image-conditioned mode can crop the subject out of the frame the user supplied. Raise it above 1 only when the user asks for more, and prefer the default when they supplied a first or last frame."
+      "description": "Strength for each LoRA in loras, in the same order. Omitting the array applies 1.0 to every LoRA, which is NOT the catalog default and for h3-realism-people is already at the top of its band, so send explicit values. Video LoRAs are positive-only — unlike the bipolar Krea 2 image sliders, a negative value is not an inverse effect and 0 is off. h3-realism-people takes 0-2 and its catalog default is 0.8; 0.6-1 is the usable band. It also pulls the camera in as it climbs: at 1.5 and above the shot reliably recomposes and the grade darkens, which on an image-conditioned mode can crop the subject out of the frame the user supplied. Raise it above 1 only when the user asks for more, and prefer the default when they supplied a first or last frame. h3-vbvr-video-reasoning and h3-mystic-xxx-v4 both take 0-1 and do default to 1.0, with usable bands of 0.7-1 and 0.2-1. h3-natural-face-speech and h3-better-motion take 0-1.5 and default to 0.6; their usable band is 0.4-0.8."
     }
   },
   "required": [

package/schemas/tools/generate_speech.schema.json

@@ -43,8 +43,7 @@
     },
     "voiceSourceIndex": {
       "type": "number",
-      "description": "Index of the uploaded audio file holding the voice to clone (0-based, from the uploaded files list, the same numbering other tools use for audio). Required when model=\"clone\" and ignored otherwise. The clip should be three to thirty seconds of one person speaking cleanly, with no music, no second speaker and no heavy room echo; anything past thirty seconds is trimmed.",
-      "minimum": 0
+      "description": "Which audio holds the voice to clone, using the same numbering every other tool uses for audio: negative indices are uploads (-1 = first/primary upload, -2 = second upload, and so on) and 0-based non-negative indices are audio generated earlier in this conversation. A clip the user just uploaded is -1. Required when model=\"clone\" and ignored otherwise. The clip should be three to thirty seconds of one person speaking cleanly, with no music, no second speaker and no heavy room echo; anything past thirty seconds is trimmed."
     },
     "voiceTranscript": {
       "type": "string",

package/schemas/tools/generate_video.schema.json

@@ -25,10 +25,6 @@
       "minimum": 2,
       "maximum": 30
     },
-    "smartDuration": {
-      "type": "boolean",
-      "description": "Wan 3 only. Set true to let Wan 3 choose an output length from 2-30 seconds. Do not also set duration. Sogni reserves the 30-second maximum before generation and settles the completed job down to Alibaba's reported output duration."
-    },
     "ratio": {
       "type": "string",
       "enum": [
@@ -77,7 +73,7 @@
         "wan3.0-video",
         "wan3.0-spicy-video"
       ],
-      "description": "\"ltx25\" (default): LTX 2.5 with native audio; Fast, HQ, and Pro currently use the release-validated official Distilled INT8 workflow. The Dev checkpoints are not publicly routed until upstream publishes and Sogni validates an official ComfyUI Dev recipe. Video model. \"ltx23\": LTX 2.3 rollback with native audio. \"wan22\": quick simple motion without audio. \"minimax-h3-t2v\": standard 20-step MiniMax H3 text-to-video; \"minimax-h3-t2v-turbo\": the existing 4-step LightX2V Turbo text-to-video; \"minimax-h3-fasth3-t2v-turbo\": the separate FastVideo VSA four-step FastH3 engine, about 2x faster and fixed to Euler/simple. All use native audio, fixed 24fps, 5.17-15.08s, and a 768p-class 32px-grid canvas; use animate_photo for H3 image-conditioned modes. Base and Turbo T2V/I2V/FLF2V prompts use the exact ordered fields integrated_multimodal_description, overall_soundscape, and non_diegetic_music; I2V/FLF2V prepend the official alignment line. \"minimax-h3-r2v\": standard 20-step MiniMax H3 reference-to-video; \"minimax-h3-r2v-turbo\": the dedicated LightX2V 4-step Ref2VA Turbo workflow using Euler/simple and a 960x544 default. FastH3 has no R2V mode. Both R2V selectors accept up to 9 images, 3 videos, and 3 audios (12 files total); at least one visual reference (image or video) is required and audio alone is invalid. Select references with referenceImageIndices/referenceVideoIndices/referenceAudioIndices and address them with the official <Subject N>/<Picture N>/<Video N>/<Audio N> semantics. Seedance quality is selected only by model: use \"seedance2-mini\" for Seedance 2.0 Mini or faster/lower-cost 720p iteration, and use \"seedance2\" for the full Seedance 2.0 model, explicit full-quality requests, 1080p/4K requests, or generated/uploaded storyboard images unless the user explicitly asks for a draft or Mini. Do not use Default Media Quality Fast/HQ/Pro or targetResolution to represent Seedance quality. \"seedance2-5\": Seedance 2.5, the newest Seedance generation — 480p and 720p ONLY (it cannot render 1080p or 4K), 4-30s per clip at a fixed 24 fps, native audio, first-and-last-frame conditioning, and a much larger reference budget than the 2.0 family: up to 30 images, 10 videos, and 10 audios, with up to 50 reference media files total, subject to those per-modality caps. Choose \"seedance2-5\" when the user asks for Seedance 2.5, wants a single continuous Seedance clip longer than 15s (2.5 renders up to 30s in one call instead of being split and stitched), or wants a first-and-last-frame Seedance transition. Keep \"seedance2\" for 1080p/4K requests, which Seedance 2.5 cannot satisfy. Seedance supports multimodal loose reference assets: images (up to 9), videos (up to 3), and audios (up to 3), with no more than 12 asset files total. Use @Image1/@Video1/@Audio1 style references in creative briefs when assigning roles. Assign every useful reference asset a role and prefer positive preservation constraints. If an uploaded video is the source clip to transform, upscale, enhance, restyle, or remaster, use video_to_video with controlMode=\"seedance-v2v\" instead of generate_video referenceVideoIndices. Alibaba HappyHorse 1.1 video models (third-party vendor — requires Premium Spark). Select by mode: \"happyhorse-1.1-t2v\" for text-to-video, \"happyhorse-1.1-i2v\" for image-to-video from one first-frame image, and \"happyhorse-1.1-r2v\" for reference-to-video with up to 9 reference images. Resolutions 720P and 1080P; duration 3-15 seconds at 24 fps; native synchronized audio is always generated (do not set generateAudio or negativePrompt). Supported aspect ratios: 16:9, 9:16, 1:1, 4:3, 3:4, 4:5, 5:4, 9:21, 21:9. HappyHorse 1.1 takes image references only and renders a native synchronized audio track (always on; do not set generateAudio or a negative prompt). Pick the model by mode: happyhorse-1.1-t2v for text-to-video (no reference image), happyhorse-1.1-i2v for image-to-video from a single first frame, and happyhorse-1.1-r2v for reference-to-video with 1 to 9 reference images. For r2v, tag the images in the prompt as [Image 1]…[Image 9] and assign each a clear role. HappyHorse does not accept reference videos or reference audios. \"wan3.0-video\" is Alibaba Wan 3 and \"wan3.0-spicy-video\" is MuleRouter w3.0-video. Both render 2-30s or smart duration at fixed 30 fps with optional native audio, provider prompt expansion, 480p/720p/1080p, adaptive/fixed ratios, first/last frames, and up to 10 image/5 video/5 audio references. Only Alibaba wan3.0-video accepts document/web context and watermark. Frame anchors and loose references are mutually exclusive. Do not send negativePrompt; video references are loose conditioning for a new result, not source-video editing or extension."
+      "description": "\"ltx25\" (default): LTX 2.5 with native audio; Fast, HQ, and Pro currently use the release-validated official Distilled INT8 workflow. The Dev checkpoints are not publicly routed until upstream publishes and Sogni validates an official ComfyUI Dev recipe. Video model. \"ltx23\": LTX 2.3 rollback with native audio. \"wan22\": quick simple motion without audio. \"minimax-h3-t2v\": standard 20-step MiniMax H3 text-to-video; \"minimax-h3-t2v-turbo\": the existing 4-step LightX2V Turbo text-to-video; \"minimax-h3-fasth3-t2v-turbo\": the separate FastVideo VSA four-step FastH3 engine, about 2x faster and fixed to Euler/simple. All use native audio, fixed 24fps, 5.17-15.08s, and a 768p-class 32px-grid canvas; use animate_photo for H3 image-conditioned modes. Base and Turbo T2V/I2V/FLF2V prompts use the exact ordered fields integrated_multimodal_description, overall_soundscape, and non_diegetic_music; I2V/FLF2V prepend the official alignment line. \"minimax-h3-r2v\": standard 20-step MiniMax H3 reference-to-video; \"minimax-h3-r2v-turbo\": the dedicated LightX2V 4-step Ref2VA Turbo workflow using Euler/simple and a 960x544 default. FastH3 has no R2V mode. Both R2V selectors accept up to 9 images, 3 videos, and 3 audios (12 files total); at least one visual reference (image or video) is required and audio alone is invalid. Select references with referenceImageIndices/referenceVideoIndices/referenceAudioIndices and address them with the official <Subject N>/<Picture N>/<Video N>/<Audio N> semantics. Seedance quality is selected only by model: use \"seedance2-mini\" for Seedance 2.0 Mini or faster/lower-cost 720p iteration, and use \"seedance2\" for the full Seedance 2.0 model, explicit full-quality requests, 1080p/4K requests, or generated/uploaded storyboard images unless the user explicitly asks for a draft or Mini. Do not use Default Media Quality Fast/HQ/Pro or targetResolution to represent Seedance quality. \"seedance2-5\": Seedance 2.5, the newest Seedance generation — 480p and 720p ONLY (it cannot render 1080p or 4K), 4-30s per clip at a fixed 24 fps, native audio, first-and-last-frame conditioning, and a much larger reference budget than the 2.0 family: up to 30 images, 10 videos, and 10 audios, with up to 50 reference media files total, subject to those per-modality caps. Choose \"seedance2-5\" when the user asks for Seedance 2.5, wants a single continuous Seedance clip longer than 15s (2.5 renders up to 30s in one call instead of being split and stitched), or wants a first-and-last-frame Seedance transition. Keep \"seedance2\" for 1080p/4K requests, which Seedance 2.5 cannot satisfy. Seedance supports multimodal loose reference assets: images (up to 9), videos (up to 3), and audios (up to 3), with no more than 12 asset files total. Use @Image1/@Video1/@Audio1 style references in creative briefs when assigning roles. Assign every useful reference asset a role and prefer positive preservation constraints. If an uploaded video is the source clip to transform, upscale, enhance, restyle, or remaster, use video_to_video with controlMode=\"seedance-v2v\" instead of generate_video referenceVideoIndices. Alibaba HappyHorse 1.1 video models (third-party vendor — requires Premium Spark). Select by mode: \"happyhorse-1.1-t2v\" for text-to-video, \"happyhorse-1.1-i2v\" for image-to-video from one first-frame image, and \"happyhorse-1.1-r2v\" for reference-to-video with up to 9 reference images. Resolutions 720P and 1080P; duration 3-15 seconds at 24 fps; native synchronized audio is always generated (do not set generateAudio or negativePrompt). Supported aspect ratios: 16:9, 9:16, 1:1, 4:3, 3:4, 4:5, 5:4, 9:21, 21:9. HappyHorse 1.1 takes image references only and renders a native synchronized audio track (always on; do not set generateAudio or a negative prompt). Pick the model by mode: happyhorse-1.1-t2v for text-to-video (no reference image), happyhorse-1.1-i2v for image-to-video from a single first frame, and happyhorse-1.1-r2v for reference-to-video with 1 to 9 reference images. For r2v, tag the images in the prompt as [Image 1]…[Image 9] and assign each a clear role. HappyHorse does not accept reference videos or reference audios. \"wan3.0-video\" is Alibaba Wan 3 and \"wan3.0-spicy-video\" is MuleRouter w3.0-video. Both render 2-30s at fixed 30 fps with optional native audio, provider prompt expansion, 480p/720p/1080p, adaptive/fixed ratios, first/last frames, and up to 10 image/5 video/5 audio references. Only Alibaba wan3.0-video accepts document/web context and watermark. Frame anchors and loose references are mutually exclusive. Do not send negativePrompt; video references are loose conditioning for a new result, not source-video editing or extension."
     },
     "generateAudio": {
       "type": "boolean",
@@ -114,7 +110,15 @@
     },
     "targetResolution": {
       "type": "number",
-      "description": "Short-side video resolution target in pixels. Use when the user asks for a bare named resolution such as \"480p\", \"720p\", \"1080p\", \"2160p\", or \"4K\" without exact pixels or an output orientation. This is resolution only, not a Seedance quality tier: Seedance quality is selected by videoModel (\"seedance2\" vs \"seedance2-mini\" vs \"seedance2-5\"). Seedance 2.0 full supports 4K; Seedance Mini and Seedance 2.5 support 480p/720p only, so never set 1080p or 4K for \"seedance2-5\". Wan 3 supports exactly 480p, 720p, and 1080p. HappyHorse supports only 720p and 1080p. Never set 4K for Wan 3 or HappyHorse. MiniMax H3 renders inside a 1344x768 pixel budget on a 32px grid, so use 768 for H3 and never 1080p or 4K. Do not set targetResolution from Default Media Quality Fast/HQ/Pro. If omitted for Seedance, Wan 3, HappyHorse, or MiniMax H3, the host uses the selected model default. This preserves/inherits the current video shape instead of forcing landscape. Do NOT set width, height, or exact-pixel aspectRatio for bare named resolution requests. If the user says \"720p portrait\", \"720p landscape\", \"4K portrait\", or \"4K landscape\", use exact width/height/aspectRatio instead."
+      "description": "Short-side video resolution target in pixels. Use when the user asks for a bare named resolution such as \"480p\", \"720p\", \"1080p\", \"2160p\", or \"4K\" without exact pixels or an output orientation. This is resolution only, not a Seedance quality tier: Seedance quality is selected by videoModel (\"seedance2\" vs \"seedance2-mini\" vs \"seedance2-5\"). Seedance 2.0 full supports 4K; Seedance Mini and Seedance 2.5 support 480p/720p only, so never set 1080p or 4K for \"seedance2-5\". Wan 3 supports exactly 480p, 720p, and 1080p. HappyHorse supports only 720p and 1080p. Never set 4K for Wan 3 or HappyHorse. MiniMax H3 renders inside a 1344x768 pixel budget on a 32px grid, so use 768 for H3 and never 1080p or 4K. Do not set targetResolution from Default Media Quality Fast/HQ/Pro. If omitted for Seedance, Wan 3, HappyHorse, or MiniMax H3, the host uses the selected model default. This preserves/inherits the current video shape instead of forcing landscape. Do NOT set width, height, or exact-pixel aspectRatio for bare named resolution requests. If the user says \"720p portrait\", \"720p landscape\", \"4K portrait\", or \"4K landscape\", use exact width/height/aspectRatio instead. For MiniMax H3 2K requests keep targetResolution at 768 (or omit it) and set outputScale to 2 instead."
+    },
+    "outputScale": {
+      "type": "integer",
+      "enum": [
+        1,
+        2
+      ],
+      "description": "MiniMax H3 only. 2 delivers 2K output: the clip renders on the normal H3 canvas and is delivered at twice its width and height (1344x768 becomes 2688x1536) with the same length and audio, for +10 Spark per second (+6 at 480p). Set 2 only when the user asks for 2K, 1440p-class or extra-sharp H3 output; leave unset otherwise. Ignored for other models."
     },
     "numberOfVariations": {
       "type": "number",
@@ -138,7 +142,7 @@
         "type": "string",
         "minLength": 1
       },
-      "description": "Ordered LoRA IDs to apply to a MiniMax H3 render. Use only when the user explicitly asks for a LoRA or for an effect one of these names describes. Stack up to 8 in one request; order matters because the adapters apply in sequence and do not commute. Keep this array positionally aligned with loraStrengths. The first render with an uncached LoRA takes longer to start while the worker downloads it.\n\nAccepted only when videoModel is one of \"minimax-h3-t2v\", \"minimax-h3-t2v-turbo\", \"minimax-h3-fasth3-t2v-turbo\", \"minimax-h3-r2v\", \"minimax-h3-r2v-turbo\". Every other video model on this tool loads no LoRAs and silently ignores these arrays, so set videoModel to an H3 mode in the same call when the user asks for one.\n\nOne LoRA is published for MiniMax H3 today: h3-realism-people (fal), a realism pass trained on live-action footage of people. It restores skin texture and pores, stray hairs, fabric weave and a fine sensor grain that the base model smooths away, and holds up in close-up. It needs its trigger word: put r34l1sm near the FRONT of the prompt, or the render comes back as ordinary H3 with no error. Exact ranges and any LoRA published since: GET /v1/loras/comfy?modelId=<model>. Do not invent ids."
+      "description": "Ordered LoRA IDs to apply to a MiniMax H3 render. Use only when the user explicitly asks for a LoRA or for an effect one of these names describes. Stack up to 8 in one request; order matters because the adapters apply in sequence and do not commute. Keep this array positionally aligned with loraStrengths. The first render with an uncached LoRA takes longer to start while the worker downloads it.\n\nAccepted only when videoModel is one of \"minimax-h3-t2v\", \"minimax-h3-t2v-turbo\", \"minimax-h3-fasth3-t2v-turbo\", \"minimax-h3-r2v\", \"minimax-h3-r2v-turbo\". Every other video model on this tool loads no LoRAs and silently ignores these arrays, so set videoModel to an H3 mode in the same call when the user asks for one.\n\nFive LoRAs are published for MiniMax H3 today and the set differs by mode, so GET /v1/loras/comfy?modelId=<model> is authoritative for the mode in hand and carries exact ranges, maturity flags, and anything published since. h3-realism-people (fal) is a realism pass trained on live-action footage of people: it restores skin texture and pores, stray hairs, fabric weave and a fine sensor grain that the base model smooths away, and holds up in close-up. It is the only one gated on a trigger word — put r34l1sm near the FRONT of the prompt, or the render comes back as ordinary H3 with no error. h3-vbvr-video-reasoning is a prompt-adherence pass that holds the model to what was asked instead of improvising. h3-natural-face-speech (AdaptiveVision) makes people talking on camera look and sound more natural: cheeks, brows, jaw and lips move together as in real speech, and spoken English comes through clearer; use it for talking-head shots such as vlogs, podcasts, interviews and presenters. h3-better-motion (AdaptiveVision) gives people more natural, consistent body movement — weight shifts, strides, turns and gestures that follow through — for dance, sport, walking and other full-body shots. Both AdaptiveVision LoRAs work best with short, simple prompt sentences and are not validated on reference-to-video. h3-mystic-xxx-v4 is an uncensored adult fine-tune. Do not invent ids."
     },
     "loraStrengths": {
       "type": "array",
@@ -147,7 +151,7 @@
       "items": {
         "type": "number"
       },
-      "description": "Strength for each LoRA in loras, in the same order. Omitting the array applies 1.0 to every LoRA, which is NOT the catalog default and for h3-realism-people is already at the top of its band, so send explicit values. Video LoRAs are positive-only — unlike the bipolar Krea 2 image sliders, a negative value is not an inverse effect and 0 is off. h3-realism-people takes 0-2 and its catalog default is 0.8; 0.6-1 is the usable band. It also pulls the camera in as it climbs: at 1.5 and above the shot reliably recomposes and the grade darkens, which on an image-conditioned mode can crop the subject out of the frame the user supplied. Raise it above 1 only when the user asks for more, and prefer the default when they supplied a first or last frame."
+      "description": "Strength for each LoRA in loras, in the same order. Omitting the array applies 1.0 to every LoRA, which is NOT the catalog default and for h3-realism-people is already at the top of its band, so send explicit values. Video LoRAs are positive-only — unlike the bipolar Krea 2 image sliders, a negative value is not an inverse effect and 0 is off. h3-realism-people takes 0-2 and its catalog default is 0.8; 0.6-1 is the usable band. It also pulls the camera in as it climbs: at 1.5 and above the shot reliably recomposes and the grade darkens, which on an image-conditioned mode can crop the subject out of the frame the user supplied. Raise it above 1 only when the user asks for more, and prefer the default when they supplied a first or last frame. h3-vbvr-video-reasoning and h3-mystic-xxx-v4 both take 0-1 and do default to 1.0, with usable bands of 0.7-1 and 0.2-1. h3-natural-face-speech and h3-better-motion take 0-1.5 and default to 0.6; their usable band is 0.4-0.8."
     }
   },
   "required": [

package/schemas/tools/sound_to_video.schema.json

@@ -3,7 +3,7 @@
   "$id": "https://schemas.sogni.ai/creative-agent/2026-07-18.1/tools/sound_to_video.schema.json",
   "title": "sound_to_video arguments",
   "schemaVersion": "2026-07-18.1",
-  "description": "Generate video synchronized to audio. Use when the user has uploaded an audio file (mp3, wav, m4a, flac) and the audio is the primary sync target, especially uploaded-audio-only workflows. Also use after generate_music (\"turn that song into a video\", \"make a music video from that\"). Auto-detects generated audio from generate_music if no audio file is uploaded. Seedance animate_photo/generate_video can also attach uploaded audio as a loose @Audio reference when an image or video reference anchors the request; use this tool instead when the soundtrack itself should drive the video. If the user provides a reference image, use ltx25-ia2v by default (ltx23-ia2v is rollback); for lip-sync with a face image, use wan-s2v; if no image, use ltx25-a2v by default (ltx23-a2v is rollback). If the user wants dialogue/audio WITHOUT pre-existing audio, use animate_photo instead (LTX 2.5 and LTX 2.3 generate audio natively). Note: Persona voice clips from resolve_personas are NOT used by this tool — for persona voice identity in video, use animate_photo or generate_video with videoModel=\"ltx23\" because LTX 2.5 has no compatible ID-LoRA. LONG AUDIO ON SEEDANCE: Seedance 2.0 and Mini cap each clip at 15s; Seedance 2.5 renders up to 30s in one call, so prefer seedance2-5 for 16-30s audio instead of splitting. When the user uploads audio longer than the per-clip cap of the selected model and Seedance is selected (seedance2, seedance2-mini, or seedance2-5), do NOT clamp to 15s and drop the rest — split the run into multiple sound_to_video calls in the same turn (one per 15s segment, so a 20s audio becomes two clips: audioStart=0 duration=15, then audioStart=15 duration=5) and finish with a single stitch_video call referencing the resulting clip indices in order with audioIndex pointing at the same uploaded audio so the stitched output carries the full original soundtrack. LTX/WAN models accept up to 20s per clip, so single-call is fine for them. Use videoModel=\"wan3.0-video\" when the user explicitly requests Wan 3 audio-driven video. Use videoModel=\"wan3.0-spicy-video\" for Wan 3.0 Enhanced audio-driven video through MuleRouter provider ID w3.0-video; it supports smart duration, adaptive ratios, and provider prompt expansion.",
+  "description": "Generate video synchronized to audio. Use when the user has uploaded an audio file (mp3, wav, m4a, flac) and the audio is the primary sync target, especially uploaded-audio-only workflows. Also use after generate_music (\"turn that song into a video\", \"make a music video from that\"). Auto-detects generated audio from generate_music if no audio file is uploaded. Seedance animate_photo/generate_video can also attach uploaded audio as a loose @Audio reference when an image or video reference anchors the request; use this tool instead when the soundtrack itself should drive the video. If the user provides a reference image, use ltx25-ia2v by default (ltx23-ia2v is rollback); for lip-sync with a face image, use wan-s2v; if no image, use ltx25-a2v by default (ltx23-a2v is rollback). If the user wants dialogue/audio WITHOUT pre-existing audio, use animate_photo instead (LTX 2.5 and LTX 2.3 generate audio natively). Note: Persona voice clips from resolve_personas are NOT used by this tool — for persona voice identity in video, use animate_photo or generate_video with videoModel=\"ltx23\" because LTX 2.5 has no compatible ID-LoRA. LONG AUDIO ON SEEDANCE: Seedance 2.0 and Mini cap each clip at 15s; Seedance 2.5 renders up to 30s in one call, so prefer seedance2-5 for 16-30s audio instead of splitting. When the user uploads audio longer than the per-clip cap of the selected model and Seedance is selected (seedance2, seedance2-mini, or seedance2-5), do NOT clamp to 15s and drop the rest — split the run into multiple sound_to_video calls in the same turn (one per 15s segment, so a 20s audio becomes two clips: audioStart=0 duration=15, then audioStart=15 duration=5) and finish with a single stitch_video call referencing the resulting clip indices in order with audioIndex pointing at the same uploaded audio so the stitched output carries the full original soundtrack. LTX/WAN models accept up to 20s per clip, so single-call is fine for them. Use videoModel=\"wan3.0-video\" when the user explicitly requests Wan 3 audio-driven video. Use videoModel=\"wan3.0-spicy-video\" for Wan 3.0 Enhanced audio-driven video through MuleRouter provider ID w3.0-video; it supports adaptive ratios and provider prompt expansion.",
   "type": "object",
   "additionalProperties": false,
   "properties": {
@@ -38,10 +38,6 @@
       "minimum": 2,
       "maximum": 30
     },
-    "smartDuration": {
-      "type": "boolean",
-      "description": "Wan 3 only. Let the model choose 2-30 seconds. Do not also set duration. Sogni reserves 30 seconds and settles down to the provider-reported duration."
-    },
     "ratio": {
       "type": "string",
       "enum": [
@@ -80,7 +76,7 @@
         "wan3.0-video",
         "wan3.0-spicy-video"
       ],
-      "description": "\"ltx25-ia2v\" (default with image) and \"ltx25-a2v\" (default without image): LTX 2.5 image+audio and audio-only modes; Fast, HQ, and Pro currently use the release-validated official Distilled INT8 workflows. The Dev checkpoints are not publicly routed until upstream publishes and Sogni validates official ComfyUI Dev recipes. Video model. \"ltx23-ia2v\" (rollback with image): LTX 2.3 image+audio to video, audio-reactive with a reference image; Fast/HQ use the distilled 8-step worker and Default Media Quality Pro uses the non-distilled dev worker. \"ltx23-a2v\" (rollback without image): LTX 2.3 audio-only to video, no image needed, creates video purely from text prompt + audio with the same quality-tier routing. \"wan-s2v\": WAN 2.2 sound-to-video, best for lip-sync with a face image, fast 4-step. \"seedance2\": full Seedance 2.0 audio-reference video, 4-15s. \"seedance2-mini\": Seedance 2.0 Mini, 720p cap, fastest/lower-cost Seedance option. Seedance quality is selected only by this model value: pick \"seedance2-mini\" for faster/lower-cost drafts or explicit Mini requests, and pick \"seedance2\" for full-quality Seedance or 1080p/4K. Do not infer the Seedance model from Default Media Quality Fast/HQ/Pro. \"seedance2-5\": Seedance 2.5, the newest Seedance generation — 480p and 720p ONLY (it cannot render 1080p or 4K), 4-30s per clip at a fixed 24 fps, native audio, first-and-last-frame conditioning, and a much larger reference budget than the 2.0 family: up to 30 images, 10 videos, and 10 audios, with up to 50 reference media files total, subject to those per-modality caps. Choose \"seedance2-5\" when the user asks for Seedance 2.5, wants a single continuous Seedance clip longer than 15s (2.5 renders up to 30s in one call instead of being split and stitched), or wants a first-and-last-frame Seedance transition. Keep \"seedance2\" for 1080p/4K requests, which Seedance 2.5 cannot satisfy. For Seedance audio-reference prompts, preserve exact spoken dialogue when the user supplied it, and assign @Image1/@Audio1 roles. If the user asks for speech without words, describe the vocal performance without inventing quoted dialogue. Treat lip-sync, voice cloning, and real-human reference behavior as provider-sensitive rather than guaranteed. Omit to auto-select based on whether an image is present. \"wan3.0-video\" is Alibaba Wan 3 and \"wan3.0-spicy-video\" is MuleRouter w3.0-video. Both render 2-30s or smart duration at fixed 30 fps with optional native audio, provider prompt expansion, 480p/720p/1080p, adaptive/fixed ratios, first/last frames, and up to 10 image/5 video/5 audio references. Only Alibaba wan3.0-video accepts document/web context and watermark. Frame anchors and loose references are mutually exclusive. Do not send negativePrompt; video references are loose conditioning for a new result, not source-video editing or extension."
+      "description": "\"ltx25-ia2v\" (default with image) and \"ltx25-a2v\" (default without image): LTX 2.5 image+audio and audio-only modes; Fast, HQ, and Pro currently use the release-validated official Distilled INT8 workflows. The Dev checkpoints are not publicly routed until upstream publishes and Sogni validates official ComfyUI Dev recipes. Video model. \"ltx23-ia2v\" (rollback with image): LTX 2.3 image+audio to video, audio-reactive with a reference image; Fast/HQ use the distilled 8-step worker and Default Media Quality Pro uses the non-distilled dev worker. \"ltx23-a2v\" (rollback without image): LTX 2.3 audio-only to video, no image needed, creates video purely from text prompt + audio with the same quality-tier routing. \"wan-s2v\": WAN 2.2 sound-to-video, best for lip-sync with a face image, fast 4-step. \"seedance2\": full Seedance 2.0 audio-reference video, 4-15s. \"seedance2-mini\": Seedance 2.0 Mini, 720p cap, fastest/lower-cost Seedance option. Seedance quality is selected only by this model value: pick \"seedance2-mini\" for faster/lower-cost drafts or explicit Mini requests, and pick \"seedance2\" for full-quality Seedance or 1080p/4K. Do not infer the Seedance model from Default Media Quality Fast/HQ/Pro. \"seedance2-5\": Seedance 2.5, the newest Seedance generation — 480p and 720p ONLY (it cannot render 1080p or 4K), 4-30s per clip at a fixed 24 fps, native audio, first-and-last-frame conditioning, and a much larger reference budget than the 2.0 family: up to 30 images, 10 videos, and 10 audios, with up to 50 reference media files total, subject to those per-modality caps. Choose \"seedance2-5\" when the user asks for Seedance 2.5, wants a single continuous Seedance clip longer than 15s (2.5 renders up to 30s in one call instead of being split and stitched), or wants a first-and-last-frame Seedance transition. Keep \"seedance2\" for 1080p/4K requests, which Seedance 2.5 cannot satisfy. For Seedance audio-reference prompts, preserve exact spoken dialogue when the user supplied it, and assign @Image1/@Audio1 roles. If the user asks for speech without words, describe the vocal performance without inventing quoted dialogue. Treat lip-sync, voice cloning, and real-human reference behavior as provider-sensitive rather than guaranteed. Omit to auto-select based on whether an image is present. \"wan3.0-video\" is Alibaba Wan 3 and \"wan3.0-spicy-video\" is MuleRouter w3.0-video. Both render 2-30s at fixed 30 fps with optional native audio, provider prompt expansion, 480p/720p/1080p, adaptive/fixed ratios, first/last frames, and up to 10 image/5 video/5 audio references. Only Alibaba wan3.0-video accepts document/web context and watermark. Frame anchors and loose references are mutually exclusive. Do not send negativePrompt; video references are loose conditioning for a new result, not source-video editing or extension."
     },
     "generateAudio": {
       "type": "boolean",

package/schemas/tools/upscale_video.schema.json

@@ -3,7 +3,7 @@
   "$id": "https://schemas.sogni.ai/creative-agent/2026-09-10.1/tools/upscale_video.schema.json",
   "title": "upscale_video arguments",
   "schemaVersion": "2026-09-10.1",
-  "description": "Upscale an existing video to 1080p or 1440p with FlashVSR video super-resolution. This is promptless, deterministic enhancement, not a generative edit: it keeps every source frame, the exact frame rate, the full aspect ratio, and the original audio, and it never changes content, trims, crops, restyles, or interpolates frames. Use it when the user asks to upscale, enlarge, sharpen, or increase the resolution of an uploaded or previously generated video, or wants an HD, 1080p, 1440p, or 2K copy of it. Do not use it to create, restyle, extend, or edit a video; when the user explicitly asks a generative model such as Seedance to re-render the clip, use video_to_video. The output is at most twice the source size, so 1080p needs a source whose short edge is 540-768px and 1440p needs 720-768px, and the source can be at most about 1344x768 pixels (768x1344 in portrait). It must also be 1-60 fps and at most 100 MB. The server sets the maximum clip length and rejects a source that is too long; never quote a length limit yourself, and if the tool reports that rejection, relay that error to the user. It cannot produce 4K or any size above 1440p; when the user asks for one, say so and offer 1440p. Each upscale costs credits based on the source video's size and length.",
+  "description": "Upscale an existing video to 1080p or 1440p with FlashVSR video super-resolution. This is promptless, deterministic enhancement, not a generative edit: it keeps every source frame, the exact frame rate, the full aspect ratio, and the original audio, and it never changes content, trims, crops, restyles, or interpolates frames. Use it when the user asks to upscale, enlarge, sharpen, or increase the resolution of an uploaded or previously generated video, or wants an HD, 1080p, 1440p, or 2K copy of it. Do not use it to create, restyle, extend, or edit a video; when the user explicitly asks a generative model such as Seedance to re-render the clip, use video_to_video. The output is at most twice the source size, so 1080p needs a source whose short edge is 540-768px and 1440p needs 720-768px, and the source can be at most about 1344x768 pixels (768x1344 in portrait). It must also be 1-60 fps and at most 100 MB. The server sets the maximum clip length and rejects a source that is too long; never quote a length limit yourself, and if the tool reports that rejection, relay that error to the user. It cannot produce 4K or any size above 1440p; when the user asks for one, say so and offer 1440p. Each upscale costs credits based on the source video's size and length. Optional detailPreference (stable or sharper), processingSpeed (stable or faster), and seed tune the result; omit them unless the user asks, and none of them changes the price.",
   "type": "object",
   "additionalProperties": false,
   "properties": {
@@ -18,6 +18,28 @@
         1440
       ],
       "description": "Output resolution of the short edge in pixels: 1440 for 1440p/2K or 1080 for 1080p/Full HD. The long edge follows the source aspect ratio, so portrait videos stay portrait. Default: 1440, or 1080 when the source's short edge is below 720px. Set 1080 when the user asks for 1080p or Full HD."
+    },
+    "detailPreference": {
+      "type": "string",
+      "enum": [
+        "stable",
+        "sharper"
+      ],
+      "description": "Detail preference. stable, the default, gives the most temporally stable result. sharper renders crisper fine detail with a little more risk of shimmer between frames. Set sharper only when the user asks for a sharper, crisper, or more detailed upscale; otherwise omit it. It does not change the price."
+    },
+    "processingSpeed": {
+      "type": "string",
+      "enum": [
+        "stable",
+        "faster"
+      ],
+      "description": "Processing speed. stable, the default, gives the most stable result. faster finishes sooner with slightly less stable detail. Set faster only when the user asks for a quicker upscale; otherwise omit it. It does not change the price."
+    },
+    "seed": {
+      "type": "integer",
+      "minimum": -1,
+      "maximum": 4294967295,
+      "description": "Seed for the upscaler's fine texture, 0 through 4294967295; different seeds give slightly different fine detail. Omit it to keep the repeatable default of 0. Use -1 for a random seed. Set it only when the user gives a seed or asks for a different or random variation of an upscale."
     }
   },
   "required": []

package/version.json

@@ -1,4 +1,4 @@
 {
-  "protocolVersion": "5.0.0",
+  "protocolVersion": "6.2.0",
   "description": "Sogni protocol artifact version. SDKs may refuse to operate against a protocolVersion they were not built for. Bump the major when removing or renaming any schema / enum / manifest field; bump the minor when adding new optional fields or new tools; bump the patch for description / prose changes only."
 }