vidspotai-shared 1.0.124 → 1.0.126

package/lib/globals/aiModels/providers/google.js

@@ -84,6 +84,49 @@ exports.googleConfigs = {
             notes: "Older Pro tier. Standard rate shown; >200k-token prompts are billed at a higher tier.",
         },
     },
+    // Gemini Omni Flash (GA 2026) — Google's current default video model. NOT a
+    // Veo model: served by the Interactions API (POST /v1beta/interactions), so
+    // google.service.ts routes it through generateOmniVideo(), not generateVideos.
+    // Docs: ai.google.dev/gemini-api/docs/omni (fetched 2026-09-29).
+    "google-gemini-omni-flash": {
+        modelId: "gemini-omni-1.1-flash",
+        type: ["text-to-video", "image-to-video"],
+        tier: "T3",
+        fields: {
+            prompt: { required: true },
+            aspectRatio: { allowedValues: ["16:9", "9:16"] },
+            // 1080p + 4k are upscaled from the native generation.
+            resolution: { allowedValues: ["360p", "720p", "1080p", "4k"] },
+            // Output is 3–10s. `response_format.duration` exists but its string
+            // format is undocumented, so the service steers length via the prompt
+            // and bills the DELIVERED length (read from the MP4 → durationSec).
+            duration: { allowedValues: [3, 4, 5, 6, 7, 8, 9, 10] },
+            imageUrl: {},
+            // Two input images = first/last-frame interpolation.
+            lastFrameImageUrl: {},
+            // Several input images = subject references (sent with task=reference_to_video).
+            referenceImageUrls: { max: 3 },
+            // No `audio` field: audio is always generated (no toggle).
+            // No negative_prompt / seed: unsupported ("put your negatives in the
+            // regular prompt") — the service folds a negative prompt into the text.
+        },
+        // Not yet published in the rate-limit table; mirror the Veo 3.1 Tier-2
+        // baseline until AI Studio shows real numbers for this key.
+        requestPerMin: 4,
+        requestPerDay: 50,
+        cost: {
+            // $17.50 / 1M output video tokens × 5,792 tokens per second of 720p video
+            // ≈ $0.10/s (ai.google.dev pricing, 2026-09-29). 1080p/4k are upscales;
+            // Google quotes no separate rate, so they bill at the same estimate.
+            perSecond: 0.10,
+        },
+        guidance: {
+            quality: 5,
+            speed: 3,
+            bestFor: "Google's newest all-round video model: text/image-to-video, first-last-frame, subject references, native audio.",
+            notes: "3–10s, 16:9 or 9:16. Audio is always on. No negative prompt or seed. 1080p/4k are upscaled.",
+        },
+    },
     "google-veo-3.1": {
         modelId: "veo-3.1-generate-preview",
         type: ["text-to-video", "image-to-video"],
@@ -155,7 +198,9 @@ exports.googleConfigs = {
         fields: {
             prompt: { required: true },
             aspectRatio: { allowedValues: ["16:9", "9:16"] },
-            resolution: { allowedValues: ["720p", "1080p"] },
+            // 4k added 2026-09-29 — the Veo parameter table gives 3.1 and 3.1 Fast
+            // the same 720p / 1080p / 4k set (1080p + 4k: 8s only; 4k not with refs).
+            resolution: { allowedValues: ["720p", "1080p", "4k"] },
             duration: { allowedValues: [4, 6, 8] },
             imageUrl: {},
             lastFrameImageUrl: {},
@@ -178,13 +223,14 @@ exports.googleConfigs = {
         durationRules: {
             4: ["720p"],
             6: ["720p"],
-            8: ["720p", "1080p"],
+            8: ["720p", "1080p", "4k"],
         },
         // Tier 2 ceiling: 4 RPM / ~50 RPD per model (see veo-3.1 note above).
         requestPerMin: 4,
         requestPerDay: 50,
         cost: {
-            perResolution: { "720p": 0.10, "1080p": 0.12 },
+            // ai.google.dev pricing (2026-09-29): $0.10 720p, $0.12 1080p, $0.30 4k.
+            perResolution: { "720p": 0.10, "1080p": 0.12, "4k": 0.30 },
         },
     },
     "google-veo-3.1-lite": {
@@ -229,7 +275,9 @@ exports.googleConfigs = {
             perResolution: { "720p": 0.05, "1080p": 0.08 },
         },
     },
-    // Deprecated by Google — migrate to veo-3.1-generate-001 before June 30, 2026
+    // ⚰️ END OF LIFE: shut down by Google 2026-06-30 (veo-3.0-generate-001,
+    // veo-3.0-fast-generate-001, veo-2.0-generate-001). Recommended replacement:
+    // veo-3.1-generate-preview / veo-3.1-fast-generate-preview (configs above).
     // "google-veo-3": {
     //   modelId: "veo-3.0-generate-001",
     //   type: ["text-to-video", "image-to-video"],
@@ -288,75 +336,173 @@ exports.googleConfigs = {
     // ───────────────────────────────────────────────────────────────────────
     // Image-gen — Imagen 4 (raw T2I) + Nano Banana (T2I + instruction edit)
     // ───────────────────────────────────────────────────────────────────────
-    // Imagen 4 — text-to-image only. No image input. aspectRatio drives output dims.
+    // ⚰️ END OF LIFE — Imagen 4 (imagen-4.0-{ultra-,,fast-}generate-001).
+    // Google's announced shutdown date was 2026-08-17 (ai.google.dev/gemini-api/
+    // docs/deprecations, recommended replacement gemini-3.1-flash-image); our
+    // live calls started 404ing 2026-08-19 and the whole family was gone from
+    // models.list by 2026-09-14. The keys are kept (saved requests, agent/MCP
+    // callers, old jobs) but REPOINTED to the closest live Gemini image model
+    // at a similar price/quality point. Fields/cost below are the successor's.
+    // REMOVED FOR v1 — kept for reference: original Imagen 4 configs.
+    // "google-imagen-4-ultra": { modelId: "imagen-4.0-ultra-generate-001", type: ["text-to-image"], tier: "T2",
+    //   fields: { prompt: { required: true }, aspectRatio: { allowedValues: ["1:1", "9:16", "16:9", "3:4", "4:3"] },
+    //     numImages: { default: 1 }, personGeneration: { allowedValues: ["allow_all", "allow_adult", "dont_allow"] } },
+    //   cost: { fixed: 0.06 } },
+    // "google-imagen-4": { modelId: "imagen-4.0-generate-001", ...same fields..., cost: { fixed: 0.04 } },
+    // "google-imagen-4-fast": { modelId: "imagen-4.0-fast-generate-001", tier: "T1", ...same fields..., cost: { fixed: 0.02 } },
+    // Imagen 4 Ultra ($0.06) → Nano Banana 2 / Gemini 3.1 Flash Image at 1K
+    // ($0.067) — Google's own recommended replacement. EOL: none announced
+    // (checked 2026-09-29).
     "google-imagen-4-ultra": {
-        modelId: "imagen-4.0-ultra-generate-001",
-        type: ["text-to-image"],
+        modelId: "gemini-3.1-flash-image",
+        deprecated: {
+            on: "2026-09-29",
+            reason: "Imagen 4 Ultra was shut down by Google on 2026-08-17; this key is served by Nano Banana 2.",
+            replacement: "google-nano-banana-2",
+        },
+        type: ["text-to-image", "image-to-image"],
         tier: "T2",
         fields: {
             prompt: { required: true },
-            aspectRatio: { allowedValues: ["1:1", "9:16", "16:9", "3:4", "4:3"] },
+            aspectRatio: { allowedValues: ["1:1", "2:3", "3:2", "3:4", "4:3", "4:5", "5:4", "9:16", "16:9", "21:9"] },
             numImages: { default: 1 },
-            personGeneration: { allowedValues: ["allow_all", "allow_adult", "dont_allow"] },
+            inputImageUrls: { max: 14 },
+        },
+        cost: { fixed: 0.067 }, // $60 / 1M image tokens, 1K output
+        guidance: {
+            quality: 4,
+            speed: 4,
+            bestFor: "Replacement for Imagen 4 Ultra — high-quality image gen + editing.",
+            notes: "Imagen 4 Ultra was shut down by Google (2026-08-17); served by gemini-3.1-flash-image.",
         },
-        cost: { fixed: 0.06 }, // per image
     },
+    // Imagen 4 ($0.04) and Imagen 4 Fast ($0.02) → Nano Banana 2 Lite / Gemini
+    // 3.1 Flash-Lite Image ($0.0336, 1K only). Cheapest live Google image model.
+    // EOL: none announced (checked 2026-09-29).
     "google-imagen-4": {
-        modelId: "imagen-4.0-generate-001",
-        type: ["text-to-image"],
-        tier: "T2",
+        modelId: "gemini-3.1-flash-lite-image",
+        deprecated: {
+            on: "2026-09-29",
+            reason: "Imagen 4 was shut down by Google on 2026-08-17; this key is served by Nano Banana 2 Lite.",
+            replacement: "google-nano-banana",
+        },
+        type: ["text-to-image", "image-to-image"],
+        tier: "T1",
         fields: {
             prompt: { required: true },
-            aspectRatio: { allowedValues: ["1:1", "9:16", "16:9", "3:4", "4:3"] },
+            aspectRatio: { allowedValues: ["1:1", "2:3", "3:2", "3:4", "4:3", "4:5", "5:4", "9:16", "16:9", "21:9"] },
             numImages: { default: 1 },
-            personGeneration: { allowedValues: ["allow_all", "allow_adult", "dont_allow"] },
+            inputImageUrls: { max: 14 },
+        },
+        cost: { fixed: 0.0336 }, // $30 / 1M image tokens, 1K output
+        guidance: {
+            quality: 3,
+            speed: 5,
+            bestFor: "Replacement for Imagen 4 — cheap, fast 1K image gen + edits.",
+            notes: "Imagen 4 was shut down by Google (2026-08-17); served by gemini-3.1-flash-lite-image.",
         },
-        cost: { fixed: 0.04 },
     },
     "google-imagen-4-fast": {
-        modelId: "imagen-4.0-fast-generate-001",
-        type: ["text-to-image"],
+        modelId: "gemini-3.1-flash-lite-image",
+        deprecated: {
+            on: "2026-09-29",
+            reason: "Imagen 4 Fast was shut down by Google on 2026-08-17; this key is served by Nano Banana 2 Lite.",
+            replacement: "google-nano-banana",
+        },
+        type: ["text-to-image", "image-to-image"],
         tier: "T1",
         fields: {
             prompt: { required: true },
-            aspectRatio: { allowedValues: ["1:1", "9:16", "16:9", "3:4", "4:3"] },
+            aspectRatio: { allowedValues: ["1:1", "2:3", "3:2", "3:4", "4:3", "4:5", "5:4", "9:16", "16:9", "21:9"] },
             numImages: { default: 1 },
-            personGeneration: { allowedValues: ["allow_all", "allow_adult", "dont_allow"] },
+            inputImageUrls: { max: 14 },
+        },
+        cost: { fixed: 0.0336 },
+        guidance: {
+            quality: 3,
+            speed: 5,
+            bestFor: "Replacement for Imagen 4 Fast — cheapest, sub-2s 1K image gen.",
+            notes: "Imagen 4 Fast was shut down by Google (2026-08-17); served by gemini-3.1-flash-lite-image.",
         },
-        cost: { fixed: 0.02 },
     },
     // Nano Banana — Gemini image models. Support T2I and instruction-edit
-    // (input images supplied via inputImageUrls; up to 14 refs for Banana 2/Pro).
+    // (input images supplied via inputImageUrls; up to 14 refs).
     "google-nano-banana-pro": {
-        modelId: "gemini-3-pro-image-preview",
+        // REMOVED FOR v1 — kept for reference: the preview id's announced shutdown
+        // was 2026-06-25 (ai.google.dev deprecations); GA replacement below.
+        // modelId: "gemini-3-pro-image-preview",
+        // EOL (gemini-3-pro-image): none announced (checked 2026-09-29).
+        modelId: "gemini-3-pro-image",
         type: ["text-to-image", "image-to-image"],
         tier: "T2",
         fields: {
             prompt: { required: true },
-            aspectRatio: { allowedValues: ["1:1", "9:16", "16:9", "3:4", "4:3"] },
+            aspectRatio: { allowedValues: ["1:1", "2:3", "3:2", "3:4", "4:3", "4:5", "5:4", "9:16", "16:9", "21:9"] },
             inputImageUrls: { max: 14 },
             imageSize: { allowedValues: ["1K", "2K", "4K", "auto_2K"] },
         },
-        // Vertex (gemini-3-pro-image-preview): output charged per-image by resolution.
+        // Output charged per-image by resolution.
         // 1K & 2K (1MP/4MP) = 1120 tokens → $0.134; 4K (16MP) = 2000 tokens → $0.24.
         cost: {
             perImageSize: { "1K": 0.134, "2K": 0.134, "auto_2K": 0.134, "4K": 0.24 },
             fixed: 0.134, // fallback when imageSize not supplied
         },
     },
-    // (Removed `google-nano-banana-2` — `gemini-2.5-pro-image` does not exist on
-    // Vertex. The 2.5 family only ships `gemini-2.5-flash-image` which is the
-    // current `google-nano-banana` entry below.)
+    // Nano Banana 2 = Gemini 3.1 Flash Image (GA 2026-05-28). EOL: none
+    // announced (checked 2026-09-29).
+    "google-nano-banana-2": {
+        modelId: "gemini-3.1-flash-image",
+        type: ["text-to-image", "image-to-image"],
+        tier: "T1",
+        fields: {
+            prompt: { required: true },
+            aspectRatio: {
+                allowedValues: [
+                    "1:1", "2:3", "3:2", "3:4", "4:3", "4:5", "5:4", "9:16", "16:9", "21:9",
+                    "1:4", "4:1", "1:8", "8:1",
+                ],
+            },
+            // Up to 14 refs total (≤10 objects hi-fi, ≤4 characters).
+            inputImageUrls: { max: 14 },
+            // "512" is the 0.5K size (uppercase K required for the others).
+            imageSize: { allowedValues: ["512", "1K", "2K", "4K"] },
+        },
+        // $60 / 1M image-output tokens (ai.google.dev pricing, 2026-09-29).
+        cost: {
+            perImageSize: { "512": 0.045, "1K": 0.067, "2K": 0.101, "4K": 0.151 },
+            fixed: 0.067, // default size is 1K
+        },
+        guidance: {
+            quality: 4,
+            speed: 4,
+            bestFor: "Fast, high-quality image gen + editing with 0.5K–4K output and extreme aspect ratios.",
+        },
+    },
     "google-nano-banana": {
-        modelId: "gemini-2.5-flash-image",
+        // ⚰️ END OF LIFE: gemini-2.5-flash-image shuts down 2026-10-02
+        // (ai.google.dev deprecations). This key is the agent/demo/MCP default
+        // image model, so it is repointed (not removed) to the closest successor
+        // at the same price point: Nano Banana 2 Lite. EOL of the successor: none
+        // announced (checked 2026-09-29).
+        // REMOVED FOR v1 — kept for reference:
+        // modelId: "gemini-2.5-flash-image",
+        // cost: { fixed: 0.0387 }, // ~$0.04 per image (1290 tokens × $30/1M)
+        modelId: "gemini-3.1-flash-lite-image",
         type: ["text-to-image", "image-to-image"],
         tier: "T1",
         fields: {
             prompt: { required: true },
-            aspectRatio: { allowedValues: ["1:1", "9:16", "16:9", "3:4", "4:3"] },
-            inputImageUrls: {},
+            aspectRatio: { allowedValues: ["1:1", "2:3", "3:2", "3:4", "4:3", "4:5", "5:4", "9:16", "16:9", "21:9"] },
+            // Up to 14 object references. Lite is 1K-only (no imageSize knob).
+            inputImageUrls: { max: 14 },
+        },
+        cost: { fixed: 0.0336 }, // $30 / 1M image tokens ≈ $0.0336 per 1K image
+        guidance: {
+            quality: 3,
+            speed: 5,
+            bestFor: "Cheapest, sub-2s image gen + local edits at 1K.",
+            notes: "Nano Banana 2 Lite. 1K only; no search grounding.",
         },
-        cost: { fixed: 0.0387 }, // ~$0.04 per image; per Google pricing (1290 tokens × $30/1M)
     },
     // ───────────────────────────────────────────────────────────────────────
     // Music-gen — Lyria 2 (Vertex AI predict endpoint, ADC auth)

package/lib/globals/aiModels/providers/kling.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"kling.d.ts","sourceRoot":"","sources":["../../../../src/globals/aiModels/providers/kling.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,WAAW,EAAE,MAAM,UAAU,CAAC;AACvC,OAAO,EAAsB,YAAY,EAAE,MAAM,UAAU,CAAC;AAqC5D,eAAO,MAAM,YAAY,EAAE,OAAO,CAAC,MAAM,CAAC,WAAW,EAAE,YAAY,CAAC,CA6bnE,CAAC"}
+{"version":3,"file":"kling.d.ts","sourceRoot":"","sources":["../../../../src/globals/aiModels/providers/kling.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,WAAW,EAAE,MAAM,UAAU,CAAC;AACvC,OAAO,EAAsB,YAAY,EAAE,MAAM,UAAU,CAAC;AAyC5D,eAAO,MAAM,YAAY,EAAE,OAAO,CAAC,MAAM,CAAC,WAAW,EAAE,YAAY,CAAC,CAsgBnE,CAAC"}

package/lib/globals/aiModels/providers/kling.js

@@ -17,7 +17,11 @@ const klingI2vFields = {
     lastFrameImageUrl: {}, // image_tail
     staticMaskUrl: {},
     dynamicMasks: {},
-    cameraControl: {},
+    // REMOVED FOR v2 — kept for reference. Kling's current 2.5-turbo / 2.6 / v3
+    // API docs (kling.ai/document-api/api/video/*, 2026-09-29) have no
+    // `camera_control`; camera moves are prompt text, which applyCameraDirection
+    // writes for every model, so no model declares a native field for it.
+    // cameraControl: {},
 };
 // ─── Kling Image Generation ───────────────────────────────────────────────────
 // Two separate endpoint families per official docs (2026-05-20):
@@ -109,7 +113,9 @@ exports.klingConfigs = {
             callbackUrl: {},
             externalTaskId: {},
             // OmniVideo uses image_list[] instead of the image/image_tail fields
-            referenceImageUrls: {}, // image_list entries without a frame type
+            // image_list entries without a frame type. Legacy omni-video docs
+            // (2026-09-29): refs + multi-image elements ≤ 7, ≤ 4 with a reference video.
+            referenceImageUrls: { max: 7 },
             inputImageUrl: {}, // image_list entry with type: first_frame
             lastFrameImageUrl: {}, // image_list entry with type: end_frame
             referenceVideoUrls: {}, // video_list[] (up to 1)
@@ -141,7 +147,7 @@ exports.klingConfigs = {
             negative_prompt: {},
             callbackUrl: {},
             externalTaskId: {},
-            referenceImageUrls: {},
+            referenceImageUrls: { max: 7 }, // see kling-video-o1
             inputImageUrl: {},
             lastFrameImageUrl: {},
             referenceVideoUrls: {},
@@ -226,10 +232,15 @@ exports.klingConfigs = {
             // values in the UI picker (P5.6.b).
             duration: { allowedValues: [3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15] },
             mode: { allowedValues: ["std", "pro", "4k"] },
-            // v3.0-only — boolean. NOT the same as v2.6's `audio` pricing-tier toggle.
-            generate_audio: {},
-            // v3.0-only — array of per-shot { prompt, duration?, startFrame? } objects.
-            // Adapter translates canonical `multiShot[]` → Kling `multi_prompt[]` + `shot_type`.
+            // Native audio → `sound: "on"|"off"` (Kling docs, 2026-09-29; the
+            // capability map lists native audio for 3.0). There is no
+            // `generate_audio` request field.
+            audio: {},
+            // REMOVED FOR v2 — kept for reference: generate_audio: {},
+            // Per-shot { prompt, duration? }. On this (text2video/image2video)
+            // endpoint Kling 3.0 reads multi-shot from the prompt — "shot n, m,
+            // words;" (docs 2026-09-29) — so buildVideoGenerateRequest encodes it
+            // there (toKlingShotPrompt). multi_prompt[] is the omni-video shape.
             multiShot: {},
         },
         concurrentRequests: 10,
@@ -274,9 +285,9 @@ exports.klingConfigs = {
     //   • no 4K (720p/1080p only)      • no first/last-frame (first_frame only)
     //   • no element control            • no motion control / video reference
     //   • no camera control             • no cfg_scale / negative_prompt fields
-    //   • multi-shot is encoded IN THE PROMPT ("shot 1, 3s, …; shot 2, 4s, …"),
-    //     not as a multi_prompt[] array — so `multiShot` is deliberately NOT
-    //     declared here; the canonical adapter would emit the wrong shape.
+    //   • multi-shot is encoded IN THE PROMPT ("shot 1, 3, …; shot 2, 4, …"),
+    //     not as a multi_prompt[] array — buildTurboVideoBody encodes the
+    //     canonical `multiShot` that way (toKlingShotPrompt, 2026-09-29).
     "kling-v3.0-turbo": {
         modelId: "kling-3.0-turbo", // also the URL path segment — keep exact
         type: ["text-to-video", "image-to-video"],
@@ -295,6 +306,7 @@ exports.klingConfigs = {
             callbackUrl: {},
             externalTaskId: {},
             watermark: {},
+            multiShot: {}, // prompt-encoded, ≤ 6 shots (see above)
         },
         concurrentRequests: 10,
         // Verified 2026-07-22 from /document-api/pricing/base/video.
@@ -327,6 +339,68 @@ exports.klingConfigs = {
             notes: "Audio is always on and always billed — for silent output kling-v3.0 std is 33% cheaper. No 4K, no last-frame, no camera control, no element/motion control.",
         },
     },
+    // ─── Kling 3.0 Omni on API 2.0 (POST /omni-video/kling-3.0-omni) ──────────
+    // The same model family as "kling-v3-omni-video" (v1 /v1/videos/omni-video,
+    // JWT auth) but on Kling's API 2.0 surface: apiKey auth, nested
+    // contents/settings/options body, data.id submit response, GET /tasks
+    // polling (shared with Turbo via checkTurboVideoStatus). What this surface
+    // adds over the v1 Omni entry:
+    //   • video EDIT via base_video (canonical inputVideoUrl)
+    //   • per-input ids the prompt cites as @image_1 / @video_1
+    //   • "original" audio (keep the edited clip's own track)
+    //
+    // Rules (enforced in buildOmni2VideoBody): last frame needs a first frame;
+    // ≤1 video; ≤7 reference images without a video, ≤4 with one; a base-video
+    // edit takes no frames and no multi-shot; a feature video allows no audio;
+    // reference/base video must be 3–15.5s. Kling Element Library inputs
+    // (`klingElements` → `element`, cited as @element_n) share the image budget
+    // (multi-image elements) or have their own caps (video-character elements);
+    // ownership is checked by KlingElementLibrary before the job is queued.
+    // Elements don't change the price tier ("video input" = feature/base video).
+    "kling-v3.0-omni": {
+        modelId: "kling-3.0-omni", // also the URL path segment — keep exact
+        type: ["text-to-video", "image-to-video", "video-to-video"],
+        tier: "T3",
+        fields: {
+            prompt: { required: true },
+            aspectRatio: { allowedValues: ["16:9", "9:16", "1:1"] },
+            duration: { allowedValues: [3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15] },
+            // std→720p, pro→1080p, 4k→4k (see resolveKlingOmni2Resolution).
+            mode: { allowedValues: ["std", "pro", "4k"] },
+            resolution: { allowedValues: ["720p", "1080p", "4k"], default: "720p" },
+            inputImageUrl: {}, // → first_frame
+            lastFrameImageUrl: {}, // → last_frame (needs a first frame)
+            referenceImageUrls: { max: 7 }, // → refer_image (≤4 with a video)
+            referenceVideoUrls: { max: 1 }, // → feature_video
+            inputVideoUrl: {}, // → base_video (edit)
+            klingElements: { max: 7 }, // → element (combination caps in buildOmni2VideoBody)
+            audio: {},
+            multiShot: {},
+            callbackUrl: {},
+            externalTaskId: {},
+            watermark: {},
+        },
+        concurrentRequests: 5,
+        // Verified 2026-09-29 from /document-api/pricing/base/video (per second):
+        //                                    720p     1080p    4k
+        //   no video input, no native audio  $0.084   $0.112   $0.42
+        //   no video input, native audio     $0.112   $0.14    $0.42
+        //   video input (never native audio) $0.126   $0.168   $0.42
+        // `perSecondByResolution` is the base (cheapest) tier so the planner's
+        // cost estimate stays conservative; getKlingCreditUsed picks the real
+        // tier from audio + inputType.
+        cost: {
+            perSecondByResolution: { "720p": 0.084, "1080p": 0.112, "4k": 0.42 },
+            nativeAudioPerSecondByResolution: { "720p": 0.112, "1080p": 0.14, "4k": 0.42 },
+            videoInputPerSecondByResolution: { "720p": 0.126, "1080p": 0.168, "4k": 0.42 },
+        },
+        guidance: {
+            quality: 5,
+            speed: 3,
+            bestFor: "Kling's most flexible 3.0 model: edit an existing clip, restyle from a reference video, or combine up to 7 reference images with first/last frames. Only Kling option with 4K plus video editing.",
+            notes: "Video input bills 50% above text/image input. A reference video forces silent output; an edit can keep the original sound. For plain text/image-to-video with sound, kling-v3.0-turbo is the same price and faster.",
+        },
+    },
     "kling-v2.6": {
         modelId: "kling-v2-6",
         type: ["text-to-video", "image-to-video"],
@@ -335,8 +409,8 @@ exports.klingConfigs = {
             ...klingCommonFields,
             ...klingI2vFields,
             mode: { allowedValues: ["std", "pro"] },
-            // v2.6 `audio` is a pricing-tier toggle, NOT Kling v3's `generate_audio`
-            // request param. Do NOT surface as a "native audio" toggle in UI on v2.x.
+            // `sound: "on"|"off"` — real native audio per Kling's capability map
+            // (2.6 "See the sound, hear the visual"), billed at the audio rate.
             audio: {},
         },
         concurrentRequests: 10,

package/lib/globals/aiModels/providers/minimax.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"minimax.d.ts","sourceRoot":"","sources":["../../../../src/globals/aiModels/providers/minimax.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,WAAW,EAAE,MAAM,UAAU,CAAC;AACvC,OAAO,EAAE,YAAY,EAAE,MAAM,UAAU,CAAC;AAExC,eAAO,MAAM,cAAc,EAAE,OAAO,CAAC,MAAM,CAAC,WAAW,EAAE,YAAY,CAAC,CAgJrE,CAAC"}
+{"version":3,"file":"minimax.d.ts","sourceRoot":"","sources":["../../../../src/globals/aiModels/providers/minimax.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,WAAW,EAAE,MAAM,UAAU,CAAC;AACvC,OAAO,EAAE,YAAY,EAAE,MAAM,UAAU,CAAC;AAKxC,eAAO,MAAM,cAAc,EAAE,OAAO,CAAC,MAAM,CAAC,WAAW,EAAE,YAAY,CAAC,CA6NrE,CAAC"}

package/lib/globals/aiModels/providers/minimax.js

@@ -1,7 +1,84 @@
 "use strict";
 Object.defineProperty(exports, "__esModule", { value: true });
 exports.minimaxConfigs = void 0;
+/** H3 `ratio` values (plus "adaptive", which we send by omitting ratio). */
+const MINIMAX_H3_ASPECTS = ["21:9", "16:9", "4:3", "1:1", "3:4", "9:16"];
 exports.minimaxConfigs = {
+    // ───────────────────────────────────────────────────────────────────────
+    // MiniMax H3 family — video generation V2 (POST /v2/video_generation,
+    // multimodal content[] with roles; poll GET /v2/query/video_generation/{id}).
+    // Source: platform.minimax.io api-reference/video-generation-v2-create +
+    // guides/pricing-paygo, read 2026-09-29. MiniMax now lists Hailuo 2.3 / 02
+    // as legacy. Pay-as-you-go only (video packages don't cover H3).
+    //   - T2V: `ratio` required. I2V (first and/or last frame): ratio follows
+    //     the image. Reference mode (≤9 images, ≤3 videos, ≤3 audio, ≤12 files
+    //     total) can't be mixed with first/last frame.
+    //   - Output always carries generated audio (dialogue, voice from a
+    //     reference_audio clip); there's no audio toggle.
+    //   - Camera: bracketed moves after key descriptions ([pan], [zoom]…).
+    // ───────────────────────────────────────────────────────────────────────
+    "minimax-h3": {
+        modelId: "MiniMax-H3",
+        type: ["text-to-video", "image-to-video", "video-to-video"],
+        tier: "T3",
+        acceptedImageFormats: ["image/jpeg", "image/jpg", "image/png", "image/webp", "image/heic", "image/heif"],
+        fields: {
+            prompt: { required: true },
+            aspectRatio: { allowedValues: MINIMAX_H3_ASPECTS },
+            resolution: { allowedValues: ["768P", "2K"], default: "768P" },
+            duration: { allowedValues: [4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15], default: 5 },
+            inputImageUrl: {},
+            lastFrameImageUrl: {},
+            referenceImageUrls: { max: 9 },
+            referenceVideoUrls: { max: 3 },
+            referenceAudioUrls: { max: 3 },
+        },
+        cost: {
+            perSecondByResolution: { "768P": 0.08, "2K": 0.13 },
+            // Reference video is billed per input second at the output-resolution rate.
+            inputVideoPerSecond: { "768P": 0.08, "2K": 0.13 },
+            freeInputImages: 5,
+            perExtraInputImage: 0.04,
+        },
+        guidance: {
+            quality: 5,
+            speed: 2,
+            bestFor: "Multimodal reference video: keep a character/product from up to 9 images, copy motion/camera from reference clips, voice from reference audio; native 2K; first+last frame.",
+            notes: "$0.40 for 5s at 768P, $0.65 at 2K. Always generates audio.",
+        },
+    },
+    "minimax-h3-max": {
+        modelId: "MiniMax-H3-Max",
+        type: ["text-to-video", "image-to-video", "video-to-video"],
+        tier: "T2",
+        acceptedImageFormats: ["image/jpeg", "image/jpg", "image/png", "image/webp", "image/heic", "image/heif"],
+        fields: {
+            prompt: { required: true },
+            aspectRatio: { allowedValues: MINIMAX_H3_ASPECTS },
+            resolution: { allowedValues: ["480P", "768P"], default: "768P" },
+            duration: { allowedValues: [5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15], default: 5 },
+            inputImageUrl: {},
+            lastFrameImageUrl: {},
+            referenceImageUrls: { max: 9 },
+            referenceVideoUrls: { max: 3 },
+            referenceAudioUrls: { max: 3 },
+            // → extra.prompt_expansion_mode: false = "disabled", else MiniMax's
+            // default "balanced".
+            promptOptimizer: {},
+        },
+        cost: {
+            perSecondByResolution: { "480P": 0.05, "768P": 0.08 },
+            inputVideoPerSecond: { "480P": 0.0553, "768P": 0.143 },
+            freeInputImages: 2,
+            perExtraInputImage: 0.074,
+        },
+        guidance: {
+            quality: 4,
+            speed: 4,
+            bestFor: "Faster H3 (fal.ai post-train) for drafts and volume: same T2V / I2V / reference modes at 480P–768P.",
+            notes: "No 2K, no 4s clips. Reference video input is pricier than on H3.",
+        },
+    },
     "minimax-hailuo-2.3-fast": {
         modelId: "MiniMax-Hailuo-2.3-Fast",
         type: ["text-to-video", "image-to-video"],

package/lib/globals/aiModels/providers/openai.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"openai.d.ts","sourceRoot":"","sources":["../../../../src/globals/aiModels/providers/openai.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,WAAW,EAAE,MAAM,UAAU,CAAC;AACvC,OAAO,EAAE,YAAY,EAAE,MAAM,UAAU,CAAC;AAExC,eAAO,MAAM,aAAa,EAAE,OAAO,CAAC,MAAM,CAAC,WAAW,EAAE,YAAY,CAAC,CAmOpE,CAAC"}
+{"version":3,"file":"openai.d.ts","sourceRoot":"","sources":["../../../../src/globals/aiModels/providers/openai.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,WAAW,EAAE,MAAM,UAAU,CAAC;AACvC,OAAO,EAAE,YAAY,EAAE,MAAM,UAAU,CAAC;AAExC,eAAO,MAAM,aAAa,EAAE,OAAO,CAAC,MAAM,CAAC,WAAW,EAAE,YAAY,CAAC,CA+OpE,CAAC"}

package/lib/globals/aiModels/providers/openai.js

@@ -5,6 +5,12 @@ exports.openaiConfigs = {
     // --- Video models ---
     "openai-sora-2": {
         modelId: "sora-2",
+        // ⚰️ END OF LIFE: OpenAI shut down Sora 2 / 2 Pro and the Videos API on
+        // 2026-09-24 (developers.openai.com deprecations).
+        retired: {
+            on: "2026-09-24",
+            reason: "OpenAI shut down Sora 2 and the Videos API on 2026-09-24.",
+        },
         type: ["text-to-video", "image-to-video", "video-to-video"],
         tier: "T3",
         fields: {
@@ -31,6 +37,12 @@ exports.openaiConfigs = {
     },
     "openai-sora-2-pro": {
         modelId: "sora-2-pro",
+        // ⚰️ END OF LIFE: OpenAI shut down Sora 2 / 2 Pro and the Videos API on
+        // 2026-09-24 (developers.openai.com deprecations).
+        retired: {
+            on: "2026-09-24",
+            reason: "OpenAI shut down Sora 2 and the Videos API on 2026-09-24.",
+        },
         type: ["text-to-video", "image-to-video", "video-to-video"],
         tier: "T3",
         fields: {

package/lib/globals/aiModels/providers/pixverse.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"pixverse.d.ts","sourceRoot":"","sources":["../../../../src/globals/aiModels/providers/pixverse.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,WAAW,EAAE,MAAM,UAAU,CAAC;AACvC,OAAO,EAAE,YAAY,EAAE,MAAM,UAAU,CAAC;AAExC,eAAO,MAAM,eAAe,EAAE,OAAO,CAAC,MAAM,CAAC,WAAW,EAAE,YAAY,CAAC,CA6LtE,CAAC"}
+{"version":3,"file":"pixverse.d.ts","sourceRoot":"","sources":["../../../../src/globals/aiModels/providers/pixverse.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,WAAW,EAAE,MAAM,UAAU,CAAC;AACvC,OAAO,EAAE,YAAY,EAAE,MAAM,UAAU,CAAC;AAsBxC,eAAO,MAAM,eAAe,EAAE,OAAO,CAAC,MAAM,CAAC,WAAW,EAAE,YAAY,CAAC,CA0PtE,CAAC"}