@picsart/ai-sdk 6.10.2 → 6.11.1

package/index.js

@@ -2866,327 +2866,623 @@ var { MODELS: MODELS8 } = defineModels("hunyuan", [
   }
 ]);
 
-// src/vendors/catalog/hailuo.ts
-var buildT2V = (withDuration) => (ctx) => ({
-  prompt: ctx.prompt,
-  ...ctx.enhancePrompt !== void 0 ? { prompt_optimizer: ctx.enhancePrompt } : {},
-  ...withDuration && ctx.duration ? { duration: ctx.duration } : {}
-});
-var buildI2V = (withDuration) => (ctx) => ({
+// src/vendors/catalog/minimax.ts
+var buildMinimaxMusicPayload = (ctx) => ({
   prompt: ctx.prompt,
-  image_url: ctx.imageUrls?.[0],
-  ...ctx.enhancePrompt !== void 0 ? { prompt_optimizer: ctx.enhancePrompt } : {},
-  ...withDuration && ctx.duration ? { duration: ctx.duration } : {}
+  ...ctx.lyricsPrompt ? { lyrics: ctx.lyricsPrompt } : {},
+  ...ctx.lyricsOptimizer != null ? { lyrics_optimizer: ctx.lyricsOptimizer } : {},
+  ...ctx.isInstrumental != null ? { is_instrumental: ctx.isInstrumental } : {},
+  audio_setting: {
+    sample_rate: ctx.sampleRate ?? 44100,
+    bitrate: ctx.bitrate ?? 256e3,
+    format: ctx.format ?? "mp3"
+  }
 });
-var buildMinimaxH3 = (ctx) => {
-  const content = [{ type: "text", text: ctx.prompt }];
-  if (ctx.startFrame) content.push({ type: "image_url", image_url: { url: ctx.startFrame }, role: "first_frame" });
-  if (ctx.endFrame) content.push({ type: "image_url", image_url: { url: ctx.endFrame }, role: "last_frame" });
-  for (const url of ctx.imageUrls ?? []) content.push({ type: "image_url", image_url: { url }, role: "reference_image" });
-  for (const url of ctx.videoUrls ?? []) content.push({ type: "video_url", video_url: { url }, role: "reference_video" });
-  for (const url of ctx.audioUrls ?? []) content.push({ type: "audio_url", audio_url: { url }, role: "reference_audio" });
-  return {
-    model: "MiniMax-H3",
-    content,
-    resolution: ctx.resolution ?? "2K",
-    duration: ctx.duration ?? 5,
-    ...ctx.aspectRatio ? { ratio: ctx.aspectRatio } : {}
-  };
+var H3_MAX_PROMPT_MAX = 5e4;
+var H3_IMAGE_BOUNDS = {
+  minSidePixels: 256,
+  maxSidePixels: 5760,
+  minAspectRatio: 0.4,
+  maxAspectRatio: 2.5
+};
+var H3_VIDEO_BOUNDS = {
+  minDurationSec: 2,
+  maxDurationSec: 15,
+  maxFrameRate: 60,
+  maxBytes: 52428800
+};
+var H3_AUDIO_BOUNDS = { minDurationSec: 2, maxDurationSec: 15 };
+var FRAME_REF_EXCLUSIVE = "References and start/end frames cannot be combined.";
+var AUDIO_NEEDS_VISUAL = "Audio cannot be the only reference \u2014 add an image or video.";
+var RATIO_FOLLOWS_FRAME = "Aspect ratio follows the frame image.";
+var ADAPTIVE_NEEDS_REFS = "Adaptive aspect ratio requires reference inputs.";
+var H3_MAX_RATIOS = ["21:9", "16:9", "4:3", "1:1", "3:4", "9:16"];
+var REFERENCE_EXCLUDES_FRAMES = {
+  startFrame: { disabled: true, reason: FRAME_REF_EXCLUSIVE },
+  endFrame: { disabled: true, reason: FRAME_REF_EXCLUSIVE }
 };
-var FRAME_REF_EXCLUSIVE = "First/last frame and reference inputs cannot be combined.";
-var LAST_NEEDS_FIRST = "An end frame requires a start frame.";
-var AUDIO_NEEDS_VISUAL = "Reference audio needs a reference image or video.";
-var minimaxH3Constraints = [
-  // Frame roles ⊥ reference roles (declared both ways so either input disables the other).
+var FRAME_EXCLUDES_REFERENCES = {
+  imageUrls: { disabled: true, reason: FRAME_REF_EXCLUSIVE },
+  videoUrls: { disabled: true, reason: FRAME_REF_EXCLUSIVE },
+  audioUrls: { disabled: true, reason: FRAME_REF_EXCLUSIVE }
+};
+var h3MaxConstraints = [
+  { when: { startFrame: { exists: true } }, then: FRAME_EXCLUDES_REFERENCES },
+  { when: { endFrame: { exists: true } }, then: FRAME_EXCLUDES_REFERENCES },
+  { when: { imageUrls: { exists: true } }, then: REFERENCE_EXCLUDES_FRAMES },
+  { when: { videoUrls: { exists: true } }, then: REFERENCE_EXCLUDES_FRAMES },
+  { when: { audioUrls: { exists: true } }, then: REFERENCE_EXCLUDES_FRAMES },
+  // With a frame the output follows that image, so the ratio is not a choice.
+  // An end frame alone is valid — it generates towards that keyframe.
   { when: { startFrame: { exists: true } }, then: {
-    imageUrls: { disabled: true, reason: FRAME_REF_EXCLUSIVE },
-    videoUrls: { disabled: true, reason: FRAME_REF_EXCLUSIVE },
-    audioUrls: { disabled: true, reason: FRAME_REF_EXCLUSIVE }
+    aspectRatio: { disabled: true, reason: RATIO_FOLLOWS_FRAME }
   } },
   { when: { endFrame: { exists: true } }, then: {
-    imageUrls: { disabled: true, reason: FRAME_REF_EXCLUSIVE },
-    videoUrls: { disabled: true, reason: FRAME_REF_EXCLUSIVE },
-    audioUrls: { disabled: true, reason: FRAME_REF_EXCLUSIVE }
-  } },
-  { when: { imageUrls: { exists: true } }, then: {
-    startFrame: { disabled: true, reason: FRAME_REF_EXCLUSIVE },
-    endFrame: { disabled: true, reason: FRAME_REF_EXCLUSIVE }
-  } },
-  { when: { videoUrls: { exists: true } }, then: {
-    startFrame: { disabled: true, reason: FRAME_REF_EXCLUSIVE },
-    endFrame: { disabled: true, reason: FRAME_REF_EXCLUSIVE }
-  } },
-  { when: { audioUrls: { exists: true } }, then: {
-    startFrame: { disabled: true, reason: FRAME_REF_EXCLUSIVE },
-    endFrame: { disabled: true, reason: FRAME_REF_EXCLUSIVE }
+    aspectRatio: { disabled: true, reason: RATIO_FOLLOWS_FRAME }
   } },
-  // last_frame requires first_frame.
-  { when: { startFrame: { exists: false } }, then: {
-    endFrame: { disabled: true, reason: LAST_NEEDS_FIRST }
+  // 'adaptive' means "follow the references" — no meaning without them.
+  { when: { imageUrls: { exists: false }, videoUrls: { exists: false }, audioUrls: { exists: false } }, then: {
+    aspectRatio: { allowed: [...H3_MAX_RATIOS], reason: ADAPTIVE_NEEDS_REFS }
   } },
-  // reference_audio cannot be the only reference input.
+  // Audio carries no visual signal, so it cannot be the only reference.
   { when: { imageUrls: { exists: false }, videoUrls: { exists: false } }, then: {
     audioUrls: { disabled: true, reason: AUDIO_NEEDS_VISUAL }
   } }
 ];
-var base = {
-  mode: "video"
-};
 var { MODELS: MODELS9 } = defineModels("minimax", [
+  // minimax-02-hd (minimax-tts) was removed in 6.0: its backend workflow was
+  // never deployed anywhere, so no generation ever existed to resolve — nothing
+  // to deprecate. Re-add as a fresh entry if MiniMax TTS ever ships.
   {
-    ...base,
-    id: "hailuo-2.3",
-    name: "Hailuo 2.3",
-    modelId: "hailuo-2.3",
-    addedAt: "2026-02-06",
-    inputType: "t2v",
-    workflow: "minimax/hailuo-2.3/standard/text-to-video",
-    editWorkflow: "minimax/hailuo-2.3/standard/image-to-video",
-    buildPayload: buildT2V(true),
-    buildEditPayload: buildI2V(true),
-    estimatedTime: 150,
-    description: "Stylized 768p animation with strong character expression and emotion.",
-    features: [feat("Image Input", "input"), feat("Start Frame", "frame"), feat("768p", "resolution"), feat("10 sec", "duration")],
-    paramConfig: {
-      ...params.prompt({ maxLength: 2e3 }),
-      ...params.enhancePrompt(),
-      ...params.duration([6, 10]),
-      ...params.imageInput()
-    }
-  },
-  {
-    ...base,
-    id: "hailuo-2.3-pro",
-    name: "Hailuo 2.3 Pro",
-    modelId: "hailuo-2.3-pro",
+    id: "minimax-music-v2",
+    name: "MiniMax Music v2",
     addedAt: "2026-02-06",
-    inputType: "t2v",
-    workflow: "minimax/hailuo-2.3/pro/text-to-video",
-    editWorkflow: "minimax/hailuo-2.3/pro/image-to-video",
-    buildPayload: buildT2V(false),
-    buildEditPayload: buildI2V(false),
-    estimatedTime: 165,
-    description: "1080p output focused on detailed scenes and polished short-form content.",
-    features: [feat("Image Input", "input"), feat("Start Frame", "frame"), feat("1080p", "resolution"), feat("6 sec", "duration")],
+    workflow: "minimax-music/v2",
+    buildPayload: buildMinimaxMusicPayload,
+    estimatedTime: 39,
+    mode: "audio",
+    inputType: "music",
+    description: "Text-to-music with vocals or instrumentals from a style prompt and lyrics prompt.",
+    features: [feat("Music", "characteristic"), feat("Vocals", "characteristic")],
     paramConfig: {
-      ...params.prompt({ maxLength: 2e3 }),
-      ...params.enhancePrompt(),
-      ...params.imageInput()
+      ...params.prompt({ maxLength: 2e3, placeholder: "Describe the genre, mood, instruments, tempo, and production style..." }),
+      ...p.text("lyricsPrompt", {
+        maxLength: 3500,
+        label: "Lyrics",
+        placeholder: "Write lyrics, or describe the lyrical theme. Optional for instrumental or optimizer-generated lyrics."
+      }),
+      ...p.boolean("lyricsOptimizer", false, "Lyrics Optimizer"),
+      ...p.boolean("isInstrumental", false, "Instrumental"),
+      ...p.enum("sampleRate", [16e3, 24e3, 32e3, 44100], 44100, { label: "Sample Rate" }),
+      ...p.enum("bitrate", [32e3, 64e3, 128e3, 256e3], 256e3, { label: "Bitrate" }),
+      ...p.enum("format", ["mp3", "wav", "pcm"], "mp3", { label: "Format" })
     }
   },
   {
-    ...base,
-    id: "hailuo-2.3-fast",
-    name: "Hailuo 2.3 Fast",
-    modelId: "hailuo-2.3-fast",
-    addedAt: "2026-02-06",
-    inputType: "i2v",
-    workflow: "minimax/hailuo-2.3-fast/standard/image-to-video",
-    buildPayload: buildI2V(true),
-    estimatedTime: 173,
-    description: "Quick 768p previews with expressive characters for rapid experimentation.",
-    features: [feat("Image Input", "input"), feat("Start Frame", "frame"), feat("768p", "resolution"), feat("10 sec", "duration")],
+    id: "minimax-music-v3",
+    name: "MiniMax Music v3",
+    addedAt: "2026-08-14",
+    workflow: "minimax-music/v3",
+    estimatedTime: 40,
+    mode: "audio",
+    inputType: "music",
+    description: "Text-to-music with vocals or instrumentals from a style prompt and optional lyrics, with configurable audio encoding.",
+    features: [feat("Music", "characteristic"), feat("Vocals", "characteristic")],
     paramConfig: {
-      ...params.prompt({ maxLength: 2e3 }),
-      ...params.enhancePrompt(),
-      ...params.duration([6, 10]),
-      ...params.imageInput(1, "Start Image", true)
+      ...params.prompt({ maxLength: 2e3, placeholder: "Describe the genre, mood, instruments, tempo, and production style..." }),
+      ...p.text("lyricsPrompt", {
+        maxLength: 3500,
+        label: "Lyrics",
+        placeholder: "Write lyrics; \\n separates lines, [Intro]/[Verse]/[Chorus] tags supported. Optional for instrumental or optimizer-generated lyrics."
+      }),
+      ...p.boolean("lyricsOptimizer", false, "Lyrics Optimizer"),
+      ...p.boolean("isInstrumental", false, "Instrumental"),
+      ...p.enum("sampleRate", [16e3, 24e3, 32e3, 44100], 44100, { label: "Sample Rate" }),
+      ...p.enum("bitrate", [32e3, 64e3, 128e3, 256e3], 256e3, { label: "Bitrate" }),
+      ...p.enum("format", ["mp3", "wav", "pcm"], "mp3", { label: "Format" })
     }
   },
   {
-    ...base,
-    id: "hailuo-2.3-fast-pro",
-    name: "Hailuo 2.3 Fast Pro",
-    modelId: "hailuo-2.3-fast-pro",
-    addedAt: "2026-02-06",
-    inputType: "i2v",
-    workflow: "minimax/hailuo-2.3-fast/pro/image-to-video",
-    buildPayload: buildI2V(false),
-    estimatedTime: 162,
-    description: "Fast 1080p output for short, polished clips with varied styles.",
-    features: [feat("Image Input", "input"), feat("Start Frame", "frame"), feat("1080p", "resolution"), feat("6 sec", "duration")],
+    // One entry for all three H3 Max modes: the inputs pick it — reference
+    // images, videos or audio generate from references; a start and/or end
+    // frame animates those frames; neither generates from the prompt alone.
+    id: "minimax-h3-max",
+    name: "MiniMax H3 Max",
+    modelId: "fal-ai-h3-max",
+    addedAt: "2026-08-28",
+    workflow: "minimax/h3-max/video-generation",
+    estimatedTime: 5,
+    mode: "video",
+    inputType: "t2v",
+    description: "Top-tier MiniMax H3 Max video from text, a start/end frame, or reference images, videos, and audio \u2014 refer to references in the prompt as Image 1, Video 1, Audio 1, in input order. Up to 15s at 1080p.",
+    features: [
+      feat("Start/End Frame", "frame"),
+      feat("Multi-Image Input", "input"),
+      feat("Video Input", "input"),
+      feat("Audio Input", "input"),
+      feat("1080p", "resolution"),
+      feat("5-15 sec", "duration")
+    ],
     paramConfig: {
-      ...params.prompt({ maxLength: 2e3 }),
-      ...params.enhancePrompt(),
-      ...params.imageInput(1, "Start Image", true)
-    }
+      ...params.prompt({ maxLength: H3_MAX_PROMPT_MAX }),
+      ...params.startFrame("Start Frame", false, H3_IMAGE_BOUNDS),
+      ...params.endFrame("End Frame", H3_IMAGE_BOUNDS),
+      // Reference slots, reference-to-video route. Per-file limits are
+      // declared; the ≤15s combined cap per modality and the ≤12 files across
+      // images + videos + audios stay backend-enforced.
+      ...params.imageInput(9, "Reference Images", false, "reference", H3_IMAGE_BOUNDS),
+      ...params.videoInputs(3, "Reference Videos", false, H3_VIDEO_BOUNDS),
+      ...params.audioInputs(3, "Reference Audios", false, H3_AUDIO_BOUNDS),
+      // 1080p is a latent refinement of a native 768p generation.
+      ...params.resolution(["480p", "768p", "1080p"], "768p"),
+      ...params.durationRange(5, 15, 5),
+      // 'adaptive' means "follow the references" — constrained below.
+      ...params.aspectRatio(["adaptive", "21:9", "16:9", "4:3", "1:1", "3:4", "9:16"], "16:9"),
+      ...p.enum("promptExpansionMode", ["balanced", "quality"], "balanced", { label: "Prompt Expansion" }),
+      // -1 (sentinel) means "pick a random seed"; the builder drops it.
+      ...p.range("seed", -1, 2147483647, -1),
+      ...p.boolean("enableSafetyChecker", true, "Safety Checker")
+    },
+    constraints: h3MaxConstraints
   },
   {
-    ...base,
-    id: "minimax-h3",
-    name: "MiniMax H3",
-    modelId: "minimax-h3",
-    addedAt: "2026-07-30",
+    // Turbo sibling of minimax-h3-max — same model family, tuned for speed
+    // (faster-than-realtime generation), text and frames only.
+    id: "minimax-h3-max-turbo",
+    name: "MiniMax H3 Max Turbo",
+    modelId: "fal-ai-h3-max-turbo",
+    addedAt: "2026-09-03",
+    workflow: "minimax/h3-max-turbo/text-to-video",
+    editWorkflow: "minimax/h3-max-turbo/image-to-video",
+    estimatedTime: 5,
+    mode: "video",
     inputType: "t2v",
-    workflow: "minimax/v2/video-generation",
-    buildPayload: buildMinimaxH3,
-    estimatedTime: 300,
-    description: "MiniMax H3 2K video from text, start/last frame, or image/video/audio references.",
+    description: "Faster-than-realtime MiniMax H3 Max Turbo video from text or a start/end frame, with prompt expansion. Up to 15s at 1080p.",
     features: [
-      feat("Start Frame", "frame"),
-      feat("End Frame", "frame"),
-      feat("Reference Video", "input"),
-      feat("2K", "resolution"),
-      feat("15 sec", "duration")
+      feat("Fast", "characteristic"),
+      feat("Image Input", "input"),
+      feat("Start/End Frame", "frame"),
+      feat("1080p", "resolution"),
+      feat("5-15 sec", "duration")
     ],
     paramConfig: {
-      ...params.prompt({ maxLength: 7e3 }),
-      ...params.startFrame(),
-      ...params.endFrame(),
-      ...params.imageInput(9, "Reference Images", false),
-      ...params.videoInputs(3, "Reference Videos", false),
-      ...params.audioInputs(3, "Reference Audios", false),
-      // Backend enum is case-insensitive (['768P','2K','768p','2k']); the
-      // uppercase forms are canonical.
-      ...params.resolution(["768P", "2K"], "2K"),
+      ...params.prompt({ maxLength: H3_MAX_PROMPT_MAX }),
+      ...params.startFrame("Start Frame", false, H3_IMAGE_BOUNDS),
+      ...params.endFrame("End Frame", H3_IMAGE_BOUNDS),
+      ...params.resolution(["480p", "768p", "1080p"], "768p"),
       ...params.durationRange(5, 15, 5),
-      ...p.aspectRatio(["adaptive", "21:9", "16:9", "4:3", "1:1", "3:4", "9:16"], "adaptive")
+      ...params.aspectRatio(["21:9", "16:9", "4:3", "1:1", "3:4", "9:16"], "16:9"),
+      ...p.enum("promptExpansionMode", ["balanced", "quality"], "balanced", { label: "Prompt Expansion" }),
+      // -1 (sentinel) means "pick a random seed"; the builder drops it.
+      ...p.range("seed", -1, 2147483647, -1),
+      ...p.boolean("enableSafetyChecker", true, "Safety Checker")
     },
-    constraints: minimaxH3Constraints
+    constraints: [
+      // Either frame routes to image-to-video, which derives the ratio from
+      // the image it was given (no aspect_ratio on that wire). An end frame
+      // alone is valid — fal documents it as end-only keyframe generation.
+      { when: { startFrame: { exists: true } }, then: {
+        aspectRatio: { disabled: true, reason: "Aspect ratio follows the frame image." }
+      } },
+      { when: { endFrame: { exists: true } }, then: {
+        aspectRatio: { disabled: true, reason: "Aspect ratio follows the frame image." }
+      } }
+    ]
+  },
+  {
+    // Camera-controls sibling of minimax-h3-max: animates a single start
+    // frame along scripted camera keyframes. The prompt is optional — blank
+    // freezes the scene and moves only the camera. No aspect ratio on this
+    // wire; the output follows the frame image.
+    id: "minimax-h3-max-camera-controls",
+    name: "MiniMax H3 Max Camera Controls",
+    modelId: "fal-ai-h3-max-camera-controls",
+    addedAt: "2026-09-14",
+    workflow: "minimax/h3-max/camera-controls",
+    estimatedTime: 5,
+    mode: "video",
+    inputType: "i2v",
+    description: "MiniMax H3 Max video from a start frame with scripted camera motion \u2014 2-12 keyframes set the camera angle, height, and distance over time. Leave the prompt blank to freeze the scene and move only the camera. Up to 15s at 1080p.",
+    features: [
+      feat("Camera Controls", "characteristic"),
+      feat("Image Input", "input"),
+      feat("1080p", "resolution"),
+      feat("5-15 sec", "duration")
+    ],
+    paramConfig: {
+      ...params.prompt({
+        maxLength: H3_MAX_PROMPT_MAX,
+        required: false,
+        placeholder: "Optional \u2014 leave blank to freeze the scene and move only the camera..."
+      }),
+      ...params.startFrame("Start Frame", true, H3_IMAGE_BOUNDS),
+      // Native 480p default here (the siblings default to 768p); 1080p is a
+      // latent refinement of a native 768p generation.
+      ...params.resolution(["480p", "768p", "1080p"], "480p"),
+      ...params.durationRange(5, 15, 5),
+      // Ordered camera keyframes: the first pose is held before its time and
+      // the final pose for the remainder. Azimuth keeps signed full turns,
+      // capped at ±32 turns of total travel; distance is in normalized scene
+      // units and must stay above zero (the wire has no upper bound).
+      cameraTrajectory: {
+        label: "Camera Trajectory",
+        descriptor: {
+          kind: "object",
+          array: { min: 2, max: 12 },
+          fields: {
+            time: { kind: "range", min: 0, max: 1, step: 0.01, default: 0 },
+            azimuth: { kind: "range", min: -11520, max: 11520, default: 0 },
+            elevation: { kind: "range", min: -90, max: 90, default: 0 },
+            distance: { kind: "range", min: 0.01, max: 100, default: 1 }
+          }
+        }
+      },
+      ...p.enum("promptExpansionMode", ["balanced", "quality"], "balanced", { label: "Prompt Expansion" }),
+      // -1 (sentinel) means "pick a random seed"; the builder drops it.
+      ...p.range("seed", -1, 2147483647, -1),
+      ...p.boolean("enableSafetyChecker", true, "Safety Checker")
+    }
+  },
+  {
+    // Lip-sync sibling of minimax-h3-max: animates a portrait to a supplied
+    // soundtrack. No prompt and no duration on this wire — the audio drives
+    // both: the vendor requires at least 5s, clips anything past 14.8s, and
+    // the output matches the clipped track. The output follows the image, so
+    // there is no aspect ratio either.
+    id: "minimax-h3-max-lip-sync",
+    name: "MiniMax H3 Max Lip Sync",
+    modelId: "fal-ai-h3-max-lip-sync",
+    addedAt: "2026-09-22",
+    workflow: "minimax/h3-max/lip-sync/image-to-video",
+    estimatedTime: 120,
+    mode: "video",
+    inputType: "i2v",
+    description: "MiniMax H3 Max lip-synced video from a portrait and an audio track \u2014 mouth movements follow the soundtrack, optionally guided by a transcript. 5-14.8s of audio, up to 2K.",
+    features: [
+      feat("Lip Sync", "characteristic"),
+      feat("Image Input", "input"),
+      feat("Audio Input", "audio"),
+      feat("2K", "resolution"),
+      feat("5-15 sec", "duration")
+    ],
+    paramConfig: {
+      ...params.startFrame("Portrait Image", true),
+      ...params.audioInput("Audio Track", true),
+      // Unlike the siblings this one also offers 2K; 1080p and 2K are latent
+      // refinements of a native 768p generation.
+      ...params.resolution(["480p", "768p", "1080p", "2k"], "768p"),
+      // Transcribes the audio to guide the sync; off follows the waveform alone.
+      ...p.boolean("enableTranscription", false, "Transcription"),
+      // -1 (sentinel) means "pick a random seed"; the builder drops it.
+      ...p.range("seed", -1, 2147483647, -1),
+      ...p.boolean("enableSafetyChecker", true, "Safety Checker")
+    }
   }
 ]);
 
-// src/vendors/catalog/wan.ts
-var buildWanT2VPayload = (ctx) => ({
-  prompt: ctx.prompt,
-  resolution: ctx.resolution ?? "720p",
-  duration: ctx.duration ?? 5,
-  ...ctx.startFrame ? { image_url: ctx.startFrame } : {},
-  ...ctx.negativePrompt ? { negative_prompt: ctx.negativePrompt } : {},
-  ...ctx.aspectRatio ? { aspect_ratio: ctx.aspectRatio } : {}
-});
-var buildWanI2VPayload = (ctx) => ({
-  prompt: ctx.prompt,
-  image_url: ctx.startFrame,
-  resolution: ctx.resolution === "1080p" ? "1080p" : "720p",
-  duration: String(ctx.duration ?? 5),
-  ...ctx.negativePrompt ? { negative_prompt: ctx.negativePrompt } : {}
-});
-var buildWanR2VPayload = (ctx) => ({
-  prompt: ctx.prompt,
-  video_urls: [ctx.videoUrl],
-  resolution: ctx.resolution === "1080p" ? "1080p" : "720p",
-  duration: String(ctx.duration ?? 5)
-});
-var buildWanImagePayload = (ctx) => ({
+// src/vendors/catalog/hailuo.ts
+var buildT2V = (withDuration) => (ctx) => ({
   prompt: ctx.prompt,
-  model: "wan-2.6",
-  count: ctx.count ?? 1,
-  ...ctx.negativePrompt ? { modelOptions: { negative_prompt: ctx.negativePrompt } } : {}
+  ...ctx.enhancePrompt !== void 0 ? { prompt_optimizer: ctx.enhancePrompt } : {},
+  ...withDuration && ctx.duration ? { duration: ctx.duration } : {}
 });
-var buildWan27T2VPayload = (ctx) => ({
+var buildI2V = (withDuration) => (ctx) => ({
   prompt: ctx.prompt,
-  resolution: ctx.resolution ?? "720P",
-  ratio: ctx.aspectRatio ?? "16:9",
-  duration: ctx.duration ?? 5,
-  prompt_extend: ctx.enhancePrompt ?? true,
-  ...ctx.negativePrompt ? { negative_prompt: ctx.negativePrompt } : {},
-  ...ctx.audioUrl ? { audio_url: ctx.audioUrl } : {},
-  ...ctx.seed != null ? { seed: ctx.seed } : {}
+  image_url: ctx.imageUrls?.[0],
+  ...ctx.enhancePrompt !== void 0 ? { prompt_optimizer: ctx.enhancePrompt } : {},
+  ...withDuration && ctx.duration ? { duration: ctx.duration } : {}
 });
-var buildWan27I2VPayload = (ctx) => {
-  const firstFrameUrl = ctx.startFrame ?? ctx.imageUrls?.[0];
-  const media = [];
-  if (firstFrameUrl) media.push({ type: "first_frame", url: firstFrameUrl });
-  if (ctx.endFrame) media.push({ type: "last_frame", url: ctx.endFrame });
-  if (ctx.audioUrl) media.push({ type: "driving_audio", url: ctx.audioUrl });
-  return {
-    media,
-    resolution: ctx.resolution ?? "720P",
-    duration: ctx.duration ?? 5,
-    prompt_extend: ctx.enhancePrompt ?? true,
-    ...ctx.prompt ? { prompt: ctx.prompt } : {},
-    ...ctx.negativePrompt ? { negative_prompt: ctx.negativePrompt } : {},
-    ...ctx.seed != null ? { seed: ctx.seed } : {}
-  };
-};
-var buildWan27R2VPayload = (ctx) => {
-  const media = [];
-  if (ctx.imageUrls?.length) {
-    for (const url of ctx.imageUrls) media.push({ type: "reference_image", url });
-  }
-  if (ctx.videoUrl) media.push({ type: "reference_video", url: ctx.videoUrl });
-  if (media.length > 5) {
-    throw new ApiError("Wan 2.7 R2V accepts at most 5 reference items (images + video combined).", {
-      status: 400,
-      code: "validation_error"
-    });
-  }
+var buildMinimaxH3 = (ctx) => {
+  const content = [{ type: "text", text: ctx.prompt }];
+  if (ctx.startFrame) content.push({ type: "image_url", image_url: { url: ctx.startFrame }, role: "first_frame" });
+  if (ctx.endFrame) content.push({ type: "image_url", image_url: { url: ctx.endFrame }, role: "last_frame" });
+  for (const url of ctx.imageUrls ?? []) content.push({ type: "image_url", image_url: { url }, role: "reference_image" });
+  for (const url of ctx.videoUrls ?? []) content.push({ type: "video_url", video_url: { url }, role: "reference_video" });
+  for (const url of ctx.audioUrls ?? []) content.push({ type: "audio_url", audio_url: { url }, role: "reference_audio" });
   return {
-    prompt: ctx.prompt,
-    media,
-    resolution: ctx.resolution ?? "720P",
-    ratio: ctx.aspectRatio ?? "16:9",
+    model: "MiniMax-H3",
+    content,
+    resolution: ctx.resolution ?? "2K",
     duration: ctx.duration ?? 5,
-    ...ctx.negativePrompt ? { negative_prompt: ctx.negativePrompt } : {},
-    ...ctx.seed != null ? { seed: ctx.seed } : {}
-  };
-};
-var buildWan27VideoEditPayload = (ctx) => {
-  const media = [];
-  if (ctx.videoUrl) media.push({ type: "video", url: ctx.videoUrl });
-  if (ctx.imageUrls?.length) {
-    for (const url of ctx.imageUrls.slice(0, 4)) media.push({ type: "reference_image", url });
-  }
-  return {
-    media,
-    resolution: ctx.resolution ?? "720P",
-    ...ctx.prompt ? { prompt: ctx.prompt } : {},
-    ...ctx.negativePrompt ? { negative_prompt: ctx.negativePrompt } : {},
-    ...ctx.aspectRatio ? { ratio: ctx.aspectRatio } : {},
-    // duration is truncation-only: unset (vendor default 0) keeps the input length.
-    ...ctx.duration ? { duration: ctx.duration } : {},
-    ...ctx.audioSetting ? { audio_setting: ctx.audioSetting } : {},
-    ...ctx.seed != null ? { seed: ctx.seed } : {}
+    ...ctx.aspectRatio ? { ratio: ctx.aspectRatio } : {}
   };
 };
-var WAN27_AR = ["16:9", "9:16", "1:1", "4:3", "3:4"];
-var WAN27_RES = ["720P", "1080P"];
-var WAN_V3_FRAME_REF_REASON = "Start/End frames cannot be combined with reference images, videos, or audios";
-var wanV3Constraints = [
-  // any reference input active → disable frame slots
+var FRAME_REF_EXCLUSIVE2 = "First/last frame and reference inputs cannot be combined.";
+var LAST_NEEDS_FIRST = "An end frame requires a start frame.";
+var AUDIO_NEEDS_VISUAL2 = "Reference audio needs a reference image or video.";
+var minimaxH3Constraints = [
+  // Frame roles ⊥ reference roles (declared both ways so either input disables the other).
+  { when: { startFrame: { exists: true } }, then: {
+    imageUrls: { disabled: true, reason: FRAME_REF_EXCLUSIVE2 },
+    videoUrls: { disabled: true, reason: FRAME_REF_EXCLUSIVE2 },
+    audioUrls: { disabled: true, reason: FRAME_REF_EXCLUSIVE2 }
+  } },
+  { when: { endFrame: { exists: true } }, then: {
+    imageUrls: { disabled: true, reason: FRAME_REF_EXCLUSIVE2 },
+    videoUrls: { disabled: true, reason: FRAME_REF_EXCLUSIVE2 },
+    audioUrls: { disabled: true, reason: FRAME_REF_EXCLUSIVE2 }
+  } },
   { when: { imageUrls: { exists: true } }, then: {
-    startFrame: { disabled: true, reason: WAN_V3_FRAME_REF_REASON },
-    endFrame: { disabled: true, reason: WAN_V3_FRAME_REF_REASON }
+    startFrame: { disabled: true, reason: FRAME_REF_EXCLUSIVE2 },
+    endFrame: { disabled: true, reason: FRAME_REF_EXCLUSIVE2 }
   } },
   { when: { videoUrls: { exists: true } }, then: {
-    startFrame: { disabled: true, reason: WAN_V3_FRAME_REF_REASON },
-    endFrame: { disabled: true, reason: WAN_V3_FRAME_REF_REASON }
+    startFrame: { disabled: true, reason: FRAME_REF_EXCLUSIVE2 },
+    endFrame: { disabled: true, reason: FRAME_REF_EXCLUSIVE2 }
   } },
   { when: { audioUrls: { exists: true } }, then: {
-    startFrame: { disabled: true, reason: WAN_V3_FRAME_REF_REASON },
-    endFrame: { disabled: true, reason: WAN_V3_FRAME_REF_REASON }
+    startFrame: { disabled: true, reason: FRAME_REF_EXCLUSIVE2 },
+    endFrame: { disabled: true, reason: FRAME_REF_EXCLUSIVE2 }
   } },
-  // any frame slot active → disable reference inputs (mirror, blocks inverse order)
-  { when: { startFrame: { exists: true } }, then: {
-    imageUrls: { disabled: true, reason: WAN_V3_FRAME_REF_REASON },
-    videoUrls: { disabled: true, reason: WAN_V3_FRAME_REF_REASON },
-    audioUrls: { disabled: true, reason: WAN_V3_FRAME_REF_REASON }
+  // last_frame requires first_frame.
+  { when: { startFrame: { exists: false } }, then: {
+    endFrame: { disabled: true, reason: LAST_NEEDS_FIRST }
   } },
-  { when: { endFrame: { exists: true } }, then: {
-    imageUrls: { disabled: true, reason: WAN_V3_FRAME_REF_REASON },
-    videoUrls: { disabled: true, reason: WAN_V3_FRAME_REF_REASON },
-    audioUrls: { disabled: true, reason: WAN_V3_FRAME_REF_REASON }
+  // reference_audio cannot be the only reference input.
+  { when: { imageUrls: { exists: false }, videoUrls: { exists: false } }, then: {
+    audioUrls: { disabled: true, reason: AUDIO_NEEDS_VISUAL2 }
   } }
 ];
-var wanV3Features = [feat("Image Input", "input"), feat("Video Input", "input"), feat("Audio", "audio"), feat("Start/End Frame", "frame"), feat("1080P", "resolution"), feat("Adaptive Ratio", "resolution")];
-var wanV3ParamConfig = {
-  // Vendor: 'either prompt or media' — the builder enforces the cross-field rule.
-  ...params.prompt({ required: false, maxLength: 5e3 }),
-  // Vendor: integer 2-30, or -1 = Smart duration mode (model picks the length).
-  duration: {
-    label: "Duration (s)",
-    descriptor: {
-      kind: "enum",
-      valueType: "number",
-      options: [{ id: -1, label: "Auto" }, { id: 5 }, { id: 10 }, { id: 15 }, { id: 30 }],
-      default: 5
-    }
-  },
-  ...params.resolution(["480P", "720P", "1080P"], "1080P"),
-  // Vendor default is 'adaptive' (model chooses from intent and input media).
-  ...params.aspectRatio(["16:9", "9:16", "1:1", "4:3", "3:4", "adaptive"], "adaptive"),
-  ...params.generateAudio(true),
+var base = {
+  mode: "video"
+};
+var { MODELS: MODELS10 } = defineModels("minimax", [
+  {
+    ...base,
+    id: "hailuo-2.3",
+    name: "Hailuo 2.3",
+    modelId: "hailuo-2.3",
+    addedAt: "2026-02-06",
+    inputType: "t2v",
+    workflow: "minimax/hailuo-2.3/standard/text-to-video",
+    editWorkflow: "minimax/hailuo-2.3/standard/image-to-video",
+    buildPayload: buildT2V(true),
+    buildEditPayload: buildI2V(true),
+    estimatedTime: 150,
+    description: "Stylized 768p animation with strong character expression and emotion.",
+    features: [feat("Image Input", "input"), feat("Start Frame", "frame"), feat("768p", "resolution"), feat("10 sec", "duration")],
+    paramConfig: {
+      ...params.prompt({ maxLength: 2e3 }),
+      ...params.enhancePrompt(),
+      ...params.duration([6, 10]),
+      ...params.imageInput()
+    }
+  },
+  {
+    ...base,
+    id: "hailuo-2.3-pro",
+    name: "Hailuo 2.3 Pro",
+    modelId: "hailuo-2.3-pro",
+    addedAt: "2026-02-06",
+    inputType: "t2v",
+    workflow: "minimax/hailuo-2.3/pro/text-to-video",
+    editWorkflow: "minimax/hailuo-2.3/pro/image-to-video",
+    buildPayload: buildT2V(false),
+    buildEditPayload: buildI2V(false),
+    estimatedTime: 165,
+    description: "1080p output focused on detailed scenes and polished short-form content.",
+    features: [feat("Image Input", "input"), feat("Start Frame", "frame"), feat("1080p", "resolution"), feat("6 sec", "duration")],
+    paramConfig: {
+      ...params.prompt({ maxLength: 2e3 }),
+      ...params.enhancePrompt(),
+      ...params.imageInput()
+    }
+  },
+  {
+    ...base,
+    id: "hailuo-2.3-fast",
+    name: "Hailuo 2.3 Fast",
+    modelId: "hailuo-2.3-fast",
+    addedAt: "2026-02-06",
+    inputType: "i2v",
+    workflow: "minimax/hailuo-2.3-fast/standard/image-to-video",
+    buildPayload: buildI2V(true),
+    estimatedTime: 173,
+    description: "Quick 768p previews with expressive characters for rapid experimentation.",
+    features: [feat("Image Input", "input"), feat("Start Frame", "frame"), feat("768p", "resolution"), feat("10 sec", "duration")],
+    paramConfig: {
+      ...params.prompt({ maxLength: 2e3 }),
+      ...params.enhancePrompt(),
+      ...params.duration([6, 10]),
+      ...params.imageInput(1, "Start Image", true)
+    }
+  },
+  {
+    ...base,
+    id: "hailuo-2.3-fast-pro",
+    name: "Hailuo 2.3 Fast Pro",
+    modelId: "hailuo-2.3-fast-pro",
+    addedAt: "2026-02-06",
+    inputType: "i2v",
+    workflow: "minimax/hailuo-2.3-fast/pro/image-to-video",
+    buildPayload: buildI2V(false),
+    estimatedTime: 162,
+    description: "Fast 1080p output for short, polished clips with varied styles.",
+    features: [feat("Image Input", "input"), feat("Start Frame", "frame"), feat("1080p", "resolution"), feat("6 sec", "duration")],
+    paramConfig: {
+      ...params.prompt({ maxLength: 2e3 }),
+      ...params.enhancePrompt(),
+      ...params.imageInput(1, "Start Image", true)
+    }
+  },
+  {
+    ...base,
+    id: "minimax-h3",
+    name: "MiniMax H3",
+    modelId: "minimax-h3",
+    addedAt: "2026-07-30",
+    inputType: "t2v",
+    workflow: "minimax/v2/video-generation",
+    buildPayload: buildMinimaxH3,
+    estimatedTime: 300,
+    description: "MiniMax H3 2K video from text, start/last frame, or image/video/audio references.",
+    features: [
+      feat("Start Frame", "frame"),
+      feat("End Frame", "frame"),
+      feat("Reference Video", "input"),
+      feat("2K", "resolution"),
+      feat("15 sec", "duration")
+    ],
+    paramConfig: {
+      ...params.prompt({ maxLength: 7e3 }),
+      // Same per-file limits as H3 Max (minimax.ts): the vendor prints the
+      // same numbers on this endpoint.
+      ...params.startFrame("Start Frame", false, H3_IMAGE_BOUNDS),
+      ...params.endFrame("End Frame", H3_IMAGE_BOUNDS),
+      ...params.imageInput(9, "Reference Images", false, "reference", H3_IMAGE_BOUNDS),
+      ...params.videoInputs(3, "Reference Videos", false, H3_VIDEO_BOUNDS),
+      ...params.audioInputs(3, "Reference Audios", false, H3_AUDIO_BOUNDS),
+      // Backend enum is case-insensitive (['768P','2K','768p','2k']); the
+      // uppercase forms are canonical.
+      ...params.resolution(["768P", "2K"], "2K"),
+      ...params.durationRange(5, 15, 5),
+      ...p.aspectRatio(["adaptive", "21:9", "16:9", "4:3", "1:1", "3:4", "9:16"], "adaptive")
+    },
+    constraints: minimaxH3Constraints
+  }
+]);
+
+// src/vendors/catalog/wan.ts
+var buildWanT2VPayload = (ctx) => ({
+  prompt: ctx.prompt,
+  resolution: ctx.resolution ?? "720p",
+  duration: ctx.duration ?? 5,
+  ...ctx.startFrame ? { image_url: ctx.startFrame } : {},
+  ...ctx.negativePrompt ? { negative_prompt: ctx.negativePrompt } : {},
+  ...ctx.aspectRatio ? { aspect_ratio: ctx.aspectRatio } : {}
+});
+var buildWanI2VPayload = (ctx) => ({
+  prompt: ctx.prompt,
+  image_url: ctx.startFrame,
+  resolution: ctx.resolution === "1080p" ? "1080p" : "720p",
+  duration: String(ctx.duration ?? 5),
+  ...ctx.negativePrompt ? { negative_prompt: ctx.negativePrompt } : {}
+});
+var buildWanR2VPayload = (ctx) => ({
+  prompt: ctx.prompt,
+  video_urls: [ctx.videoUrl],
+  resolution: ctx.resolution === "1080p" ? "1080p" : "720p",
+  duration: String(ctx.duration ?? 5)
+});
+var buildWanImagePayload = (ctx) => ({
+  prompt: ctx.prompt,
+  model: "wan-2.6",
+  count: ctx.count ?? 1,
+  ...ctx.negativePrompt ? { modelOptions: { negative_prompt: ctx.negativePrompt } } : {}
+});
+var buildWan27T2VPayload = (ctx) => ({
+  prompt: ctx.prompt,
+  resolution: ctx.resolution ?? "720P",
+  ratio: ctx.aspectRatio ?? "16:9",
+  duration: ctx.duration ?? 5,
+  prompt_extend: ctx.enhancePrompt ?? true,
+  ...ctx.negativePrompt ? { negative_prompt: ctx.negativePrompt } : {},
+  ...ctx.audioUrl ? { audio_url: ctx.audioUrl } : {},
+  ...ctx.seed != null ? { seed: ctx.seed } : {}
+});
+var buildWan27I2VPayload = (ctx) => {
+  const firstFrameUrl = ctx.startFrame ?? ctx.imageUrls?.[0];
+  const media = [];
+  if (firstFrameUrl) media.push({ type: "first_frame", url: firstFrameUrl });
+  if (ctx.endFrame) media.push({ type: "last_frame", url: ctx.endFrame });
+  if (ctx.audioUrl) media.push({ type: "driving_audio", url: ctx.audioUrl });
+  return {
+    media,
+    resolution: ctx.resolution ?? "720P",
+    duration: ctx.duration ?? 5,
+    prompt_extend: ctx.enhancePrompt ?? true,
+    ...ctx.prompt ? { prompt: ctx.prompt } : {},
+    ...ctx.negativePrompt ? { negative_prompt: ctx.negativePrompt } : {},
+    ...ctx.seed != null ? { seed: ctx.seed } : {}
+  };
+};
+var buildWan27R2VPayload = (ctx) => {
+  const media = [];
+  if (ctx.imageUrls?.length) {
+    for (const url of ctx.imageUrls) media.push({ type: "reference_image", url });
+  }
+  if (ctx.videoUrl) media.push({ type: "reference_video", url: ctx.videoUrl });
+  if (media.length > 5) {
+    throw new ApiError("Wan 2.7 R2V accepts at most 5 reference items (images + video combined).", {
+      status: 400,
+      code: "validation_error"
+    });
+  }
+  return {
+    prompt: ctx.prompt,
+    media,
+    resolution: ctx.resolution ?? "720P",
+    ratio: ctx.aspectRatio ?? "16:9",
+    duration: ctx.duration ?? 5,
+    ...ctx.negativePrompt ? { negative_prompt: ctx.negativePrompt } : {},
+    ...ctx.seed != null ? { seed: ctx.seed } : {}
+  };
+};
+var buildWan27VideoEditPayload = (ctx) => {
+  const media = [];
+  if (ctx.videoUrl) media.push({ type: "video", url: ctx.videoUrl });
+  if (ctx.imageUrls?.length) {
+    for (const url of ctx.imageUrls.slice(0, 4)) media.push({ type: "reference_image", url });
+  }
+  return {
+    media,
+    resolution: ctx.resolution ?? "720P",
+    ...ctx.prompt ? { prompt: ctx.prompt } : {},
+    ...ctx.negativePrompt ? { negative_prompt: ctx.negativePrompt } : {},
+    ...ctx.aspectRatio ? { ratio: ctx.aspectRatio } : {},
+    // duration is truncation-only: unset (vendor default 0) keeps the input length.
+    ...ctx.duration ? { duration: ctx.duration } : {},
+    ...ctx.audioSetting ? { audio_setting: ctx.audioSetting } : {},
+    ...ctx.seed != null ? { seed: ctx.seed } : {}
+  };
+};
+var WAN27_AR = ["16:9", "9:16", "1:1", "4:3", "3:4"];
+var WAN27_RES = ["720P", "1080P"];
+var WAN_V3_FRAME_REF_REASON = "Start/End frames cannot be combined with reference images, videos, or audios";
+var wanV3Constraints = [
+  // any reference input active → disable frame slots
+  { when: { imageUrls: { exists: true } }, then: {
+    startFrame: { disabled: true, reason: WAN_V3_FRAME_REF_REASON },
+    endFrame: { disabled: true, reason: WAN_V3_FRAME_REF_REASON }
+  } },
+  { when: { videoUrls: { exists: true } }, then: {
+    startFrame: { disabled: true, reason: WAN_V3_FRAME_REF_REASON },
+    endFrame: { disabled: true, reason: WAN_V3_FRAME_REF_REASON }
+  } },
+  { when: { audioUrls: { exists: true } }, then: {
+    startFrame: { disabled: true, reason: WAN_V3_FRAME_REF_REASON },
+    endFrame: { disabled: true, reason: WAN_V3_FRAME_REF_REASON }
+  } },
+  // any frame slot active → disable reference inputs (mirror, blocks inverse order)
+  { when: { startFrame: { exists: true } }, then: {
+    imageUrls: { disabled: true, reason: WAN_V3_FRAME_REF_REASON },
+    videoUrls: { disabled: true, reason: WAN_V3_FRAME_REF_REASON },
+    audioUrls: { disabled: true, reason: WAN_V3_FRAME_REF_REASON }
+  } },
+  { when: { endFrame: { exists: true } }, then: {
+    imageUrls: { disabled: true, reason: WAN_V3_FRAME_REF_REASON },
+    videoUrls: { disabled: true, reason: WAN_V3_FRAME_REF_REASON },
+    audioUrls: { disabled: true, reason: WAN_V3_FRAME_REF_REASON }
+  } }
+];
+var wanV3Features = [feat("Image Input", "input"), feat("Video Input", "input"), feat("Audio", "audio"), feat("Start/End Frame", "frame"), feat("1080P", "resolution"), feat("Adaptive Ratio", "resolution")];
+var wanV3ParamConfig = {
+  // Vendor: 'either prompt or media' — the builder enforces the cross-field rule.
+  ...params.prompt({ required: false, maxLength: 5e3 }),
+  // Vendor: integer 2-30, or -1 = Smart duration mode (model picks the length).
+  duration: {
+    label: "Duration (s)",
+    descriptor: {
+      kind: "enum",
+      valueType: "number",
+      options: [{ id: -1, label: "Auto" }, { id: 5 }, { id: 10 }, { id: 15 }, { id: 30 }],
+      default: 5
+    }
+  },
+  ...params.resolution(["480P", "720P", "1080P"], "1080P"),
+  // Vendor default is 'adaptive' (model chooses from intent and input media).
+  ...params.aspectRatio(["16:9", "9:16", "1:1", "4:3", "3:4", "adaptive"], "adaptive"),
+  ...params.generateAudio(true),
   ...params.startFrame(),
   ...params.endFrame(),
   ...params.imageInput(10, "Reference Images"),
@@ -3197,7 +3493,7 @@ var wanV3ParamConfig = {
   // No default: a materialized seed would pin every generation to one value.
   ...params.seed()
 };
-var { MODELS: MODELS10 } = defineModels("wan", [
+var { MODELS: MODELS11 } = defineModels("wan", [
   // ── Video ─────────────────────────────────────────
   {
     id: "wan-2.6-t2v",
@@ -3448,7 +3744,7 @@ var makeWanV3VideoPayload = (model) => (input) => {
     ...input.seed != null ? { seed: input.seed } : {}
   };
 };
-registerPayloads(MODELS10, {
+registerPayloads(MODELS11, {
   "wan-3.0-video": makeWanV3VideoPayload("wan3.0-video"),
   "wan-3.0-video-prime": makeWanV3VideoPayload("wan3.0-video-prime")
 });
@@ -3597,7 +3893,7 @@ var ray32EditConstraints = [
     exrExport: { disabled: true, reason: HDR_540P }
   } }
 ];
-var { MODELS: MODELS11 } = defineModels("luma", [
+var { MODELS: MODELS12 } = defineModels("luma", [
   {
     id: "luma-ray-2",
     name: "Luma Ray 2",
@@ -3807,7 +4103,7 @@ var buildRay32ReframePayload = (input) => ({
     resolution: input.resolution ?? "720p"
   }
 });
-registerPayloads(MODELS11, {
+registerPayloads(MODELS12, {
   "luma-ray-3.2": buildRay32VideoPayload,
   "luma-ray-3.2-edit": buildRay32EditPayload,
   "luma-ray-3.2-reframe-video": buildRay32ReframePayload
@@ -4196,7 +4492,7 @@ var seedance25DraftTaskParam = {
     }
   }
 };
-var { MODELS: MODELS12 } = defineModels("seedance", [
+var { MODELS: MODELS13 } = defineModels("seedance", [
   {
     id: "seedance-2.5",
     name: "Seedance 2.5",
@@ -4781,7 +5077,7 @@ var seedreamV2Params = {
   ...params.imageInput(2, "Source Images"),
   ...params.negativePrompt()
 };
-var { MODELS: MODELS13 } = defineModels("seedream", [
+var { MODELS: MODELS14 } = defineModels("seedream", [
   {
     id: "seedream-5.0-pro",
     name: "Seedream 5.0 Pro",
@@ -4925,7 +5221,7 @@ var refMutexConstraints = [
     imageUrls: { disabled: true, reason: REF_MUTEX_REASON }
   } }
 ];
-var { MODELS: MODELS14 } = defineModels("seedaudio", [
+var { MODELS: MODELS15 } = defineModels("seedaudio", [
   {
     id: "seed-audio-1.0-multilingual",
     name: "Seed Audio Multilingual",
@@ -4991,7 +5287,7 @@ var buildSeedAudioPayload = (model) => (input) => {
     ...references ? { references } : {}
   };
 };
-registerPayloads(MODELS14, {
+registerPayloads(MODELS15, {
   "seed-audio-1.0": buildSeedAudioPayload("seed-audio-1.0"),
   "seed-audio-1.0-multilingual": buildSeedAudioPayload("seed-audio-1.0-multilingual")
 });
@@ -5007,7 +5303,7 @@ var buildRevePayload = (ctx) => {
     ...hasImages ? { image_url: ctx.imageUrls[0] } : {}
   };
 };
-var { MODELS: MODELS15 } = defineModels("reve", [
+var { MODELS: MODELS16 } = defineModels("reve", [
   {
     id: "reve",
     name: "Reve",
@@ -5103,7 +5399,7 @@ var GROK_DURATIONS = [3, 5, 6, 8, 10, 12, 15];
 var GROK_VIDEO_RESOLUTIONS = ["480p", "720p"];
 var GROK_VIDEO_RESOLUTIONS_15 = ["480p", "720p", "1080p"];
 var GROK_IMAGE_RESOLUTIONS = ["1k", "2k"];
-var { MODELS: MODELS16 } = defineModels("grok", [
+var { MODELS: MODELS17 } = defineModels("grok", [
   // ── Video ─────────────────────────────────────────
   {
     id: "grok-imagine-video",
@@ -5292,10 +5588,10 @@ var buildGrokImage2EditPayload = (input) => {
     ...imagePart2
   };
 };
-registerPayloads(MODELS16, {
+registerPayloads(MODELS17, {
   "grok-imagine-image-2.0": buildGrokImage2Payload
 });
-registerEditPayloads(MODELS16, {
+registerEditPayloads(MODELS17, {
   "grok-imagine-image-2.0": buildGrokImage2EditPayload
 });
 
@@ -5333,7 +5629,7 @@ var buildPikaFramesPayload = (ctx) => ({
 var PIKA_DURATIONS = [5, 10];
 var PIKA_AR = ["16:9", "9:16", "1:1", "4:5", "5:4", "3:2", "2:3"];
 var PIKA_RESOLUTIONS = ["720p", "1080p"];
-var { MODELS: MODELS17 } = defineModels("pika", [
+var { MODELS: MODELS18 } = defineModels("pika", [
   {
     id: "pika-2.2",
     name: "Pika",
@@ -5462,7 +5758,7 @@ var veoLiteParamConfig = {
 var veoLiteConstraints = [
   { when: { resolution: { is: "1080p" } }, then: { duration: { allowed: [8], reason: "1080p supports 8s only" } } }
 ];
-var { MODELS: MODELS18 } = defineModels("google", [
+var { MODELS: MODELS19 } = defineModels("google", [
   {
     id: "veo-3.1",
     name: "Veo 3.1",
@@ -5598,7 +5894,7 @@ var buildRunwayGen3aTurboPayload = (ctx) => ({
   ratio: RUNWAY_GEN3A_RATIO_MAP[ctx.aspectRatio ?? "16:9"] ?? "1280:768",
   duration: ctx.duration ?? 5
 });
-var { MODELS: MODELS19 } = defineModels("runway", [
+var { MODELS: MODELS20 } = defineModels("runway", [
   {
     id: "runway-avatar-video",
     name: "Runway Avatar",
@@ -5793,7 +6089,7 @@ var flux3VideoConstraints = [
     resolution: { allowed: ["hd"], reason: "Draft mode only supports HD resolution." }
   } }
 ];
-var { MODELS: MODELS20 } = defineModels("flux", [
+var { MODELS: MODELS21 } = defineModels("flux", [
   {
     ...fluxV2Base,
     id: "flux-2-pro",
@@ -6042,7 +6338,7 @@ var buildFlux3VideoPayload = (input) => {
     ...input.draft ? { draft: input.draft } : {}
   };
 };
-registerPayloads(MODELS20, {
+registerPayloads(MODELS21, {
   "flux-3-video": buildFlux3VideoPayload
 });
 
@@ -6124,7 +6420,7 @@ var thinkingBudgetParam = {
     descriptor: { kind: "range", min: 128, max: 24576, step: 128, default: 128 }
   }
 };
-var { MODELS: MODELS21 } = defineModels("google", [
+var { MODELS: MODELS22 } = defineModels("google", [
   // ── Image ─────────────────────────────────────────────────────────
   {
     id: "gemini-3.1-flash-image",
@@ -6322,7 +6618,7 @@ var buildOmniFlash11Payload = (input) => ({
   ...input.videoUrl ? { video: { url: input.videoUrl } } : {},
   ...input.videoUrls?.length ? { referenceVideos: input.videoUrls.map((url) => ({ url })) } : {}
 });
-registerPayloads(MODELS21, {
+registerPayloads(MODELS22, {
   "gemini-omni-1.1-flash-preview": buildOmniFlash11Payload
 });
 
@@ -6428,7 +6724,7 @@ var gptImageBgConstraints = [
     then: { outputFormat: { allowed: ["png", "webp"], reason: "Transparent background needs PNG or WEBP \u2014 JPEG has no alpha channel." } }
   }
 ];
-var { MODELS: MODELS22 } = defineModels("openai", [
+var { MODELS: MODELS23 } = defineModels("openai", [
   // ── Image ─────────────────────────────────────────
   {
     id: "gpt-image-2.5-sunburst",
@@ -6575,11 +6871,11 @@ var editFor = (model) => (input) => ({
   ...input.background ? { background: input.background } : {},
   ...input.outputFormat ? { output_format: input.outputFormat } : {}
 });
-registerPayloads(MODELS22, {
+registerPayloads(MODELS23, {
   "gpt-image-2.5-flare": generateFor("gpt-image-2.5-flare"),
   "gpt-image-2.5-sunburst": generateFor("gpt-image-2.5-sunburst")
 });
-registerEditPayloads(MODELS22, {
+registerEditPayloads(MODELS23, {
   "gpt-image-2.5-flare": editFor("gpt-image-2.5-flare"),
   "gpt-image-2.5-sunburst": editFor("gpt-image-2.5-sunburst")
 });
@@ -6643,7 +6939,7 @@ var ttsParamConfig = (promptMaxLength, withLanguage) => ({
   ...params.voiceId([], DEFAULT_VOICE_ID, { catalog: VOICE_CATALOG }),
   ...voiceSettingsParams(true)
 });
-var { MODELS: MODELS23 } = defineModels("elevenlabs", [
+var { MODELS: MODELS24 } = defineModels("elevenlabs", [
   // ── TTS ───────────────────────────────────────────────────────────
   {
     id: "eleven-v3",
@@ -6859,480 +7155,199 @@ var { MODELS: MODELS23 } = defineModels("elevenlabs", [
         label: "Voice ID (a custom voice from your workspace)",
         required: true,
         descriptor: { kind: "text", placeholder: "Premade/catalog voices are rejected by ElevenLabs" }
-      },
-      ...params.prompt({ minLength: 5, maxLength: 1e3 })
-    }
-  },
-  {
-    id: "eleven-voice-design-v3",
-    name: "Eleven Voice Design v3",
-    addedAt: "2026-03-24",
-    modelId: "eleven_ttv_v3",
-    workflow: "elevenlabs/v1/voice-design",
-    buildPayload: buildElevenLabsVoiceDesignPayload("eleven_ttv_v3"),
-    estimatedTime: 15,
-    mode: "audio",
-    inputType: "tts",
-    description: "Design a new voice from a text description using v3 engine.",
-    features: [feat("Voice Design", "characteristic"), feat("Preview", "characteristic")],
-    paramConfig: { ...params.prompt({ minLength: 20, maxLength: 1e3 }) }
-  },
-  {
-    id: "eleven-voice-design-v2",
-    name: "Eleven Voice Design Multilingual v2",
-    addedAt: "2026-03-24",
-    modelId: "eleven_multilingual_ttv_v2",
-    workflow: "elevenlabs/v1/voice-design",
-    buildPayload: buildElevenLabsVoiceDesignPayload("eleven_multilingual_ttv_v2"),
-    estimatedTime: 15,
-    mode: "audio",
-    inputType: "tts",
-    description: "Design a new voice from a text description with multilingual support.",
-    features: [feat("Voice Design", "characteristic"), feat("Multilingual", "characteristic"), feat("Preview", "characteristic")],
-    paramConfig: { ...params.prompt({ minLength: 20, maxLength: 1e3 }) }
-  },
-  {
-    id: "eleven-voice-create",
-    name: "Eleven Voice Previews",
-    // Pricing registers this operation as `eleven-voice-create-previews`, which
-    // does not match the model id — without the hint the credits lookup misses.
-    modelId: "eleven-voice-create-previews",
-    addedAt: "2026-03-24",
-    workflow: "elevenlabs/v1/voice-create-previews",
-    buildPayload: buildElevenLabsVoicePreviewsPayload,
-    estimatedTime: 15,
-    mode: "audio",
-    inputType: "tts",
-    description: "Generate voice previews from a description to audition before committing.",
-    features: [feat("Voice Design", "characteristic"), feat("Preview", "characteristic")],
-    paramConfig: { ...params.prompt({ minLength: 20, maxLength: 1e3 }) }
-  }
-]);
-
-// src/vendors/catalog/elevenlabs.payloads.ts
-var buildElevenLabsMusicPayload = (input) => ({
-  prompt: input.prompt,
-  music_length_seconds: input.duration ?? 30,
-  model_id: "music_v2",
-  force_instrumental: input.isInstrumental ?? false
-});
-var voiceSettings = (input) => {
-  const settings = {
-    ...input.stability != null ? { stability: input.stability } : {},
-    ...input.similarityBoost != null ? { similarity_boost: input.similarityBoost } : {},
-    ...input.styleExaggeration != null ? { style: input.styleExaggeration } : {},
-    ...input.speed != null ? { speed: input.speed } : {},
-    ...input.useSpeakerBoost != null ? { use_speaker_boost: input.useSpeakerBoost } : {}
-  };
-  return Object.keys(settings).length > 0 ? { voice_settings: settings } : {};
-};
-var buildElevenLabsTTSPayload = (modelId) => (input) => ({
-  text: input.prompt,
-  voice_id: input.voiceId ?? DEFAULT_VOICE_ID,
-  model_id: modelId,
-  ...input.language ? { language_code: input.language } : {},
-  ...voiceSettings(input)
-});
-var buildElevenLabsSTSPayload = (modelId) => (input) => ({
-  audio_url: input.audioUrl,
-  voice_id: input.voiceId ?? DEFAULT_VOICE_ID,
-  model_id: modelId,
-  remove_background_noise: input.removeBackgroundNoise ?? false,
-  ...voiceSettings(input)
-});
-var buildElevenLabsDialoguePayload = (input) => ({
-  conversation: input.dialogue.map((line) => ({
-    voice_id: line.voiceId,
-    text: line.text
-  })),
-  model_id: "eleven_v3",
-  ...input.stability != null ? { settings: { stability: input.stability } } : {},
-  ...input.language ? { language_code: input.language } : {},
-  ...input.seed != null ? { seed: input.seed } : {}
-});
-registerPayloads(MODELS23, {
-  "elevenlabs-music-v2": buildElevenLabsMusicPayload,
-  "eleven-v3": buildElevenLabsTTSPayload("eleven_v3"),
-  "eleven-multilingual-v2": buildElevenLabsTTSPayload("eleven_multilingual_v2"),
-  "eleven-sts-v2": buildElevenLabsSTSPayload("eleven_english_sts_v2"),
-  "eleven-multilingual-sts-v2": buildElevenLabsSTSPayload("eleven_multilingual_sts_v2"),
-  "eleven-text-to-dialogue": buildElevenLabsDialoguePayload
-});
-
-// src/vendors/catalog/heygen.ts
-var buildHeyGenPhotoAvatarPayload = (ctx) => ({
-  image_url: ctx.imageUrls?.[0],
-  script: ctx.prompt,
-  voice_id: ctx.voiceId || void 0,
-  ...ctx.resolution ? { resolution: ctx.resolution } : {},
-  ...ctx.aspectRatio ? { aspect_ratio: ctx.aspectRatio } : {}
-});
-var buildHeyGenVideoAvatarPayload = (ctx) => {
-  const c = ctx;
-  return {
-    avatar_id: ctx.videoId || void 0,
-    script: ctx.prompt,
-    voice_id: ctx.voiceId || void 0,
-    ...ctx.resolution ? { resolution: ctx.resolution } : {},
-    ...ctx.aspectRatio ? { aspect_ratio: ctx.aspectRatio } : {},
-    ...c.engine ? { engine: c.engine } : {}
-  };
-};
-var engineParam = {
-  engine: {
-    label: "Engine",
-    required: false,
-    descriptor: {
-      kind: "enum",
-      valueType: "string",
-      options: [
-        { id: "avatar_iv", label: "Avatar IV" },
-        { id: "avatar_v", label: "Avatar V" }
-      ],
-      default: "avatar_iv"
-    }
-  }
-};
-var dynamicVoiceConfig = {
-  ...params.voiceId([], "", {
-    required: true,
-    catalog: { workflow: "heygen/v1/catalog/voices" }
-  })
-};
-var { MODELS: MODELS24 } = defineModels("heygen", [
-  // ── Photo Avatar (i2v) ────────────────────────────────────────────
-  {
-    id: "heygen-talking-photo",
-    modelId: "heygen-avatar-iv",
-    addedAt: "2026-03-17",
-    name: "HeyGen Talking Photo",
-    workflow: "heygen/v1/video/generate",
-    buildPayload: buildHeyGenPhotoAvatarPayload,
-    estimatedTime: 120,
-    mode: "video",
-    inputType: "i2v",
-    description: "Animate any photo into a speaking avatar with natural lip-sync.",
-    features: [
-      feat("Image Input", "input"),
-      feat("Voice Selection", "audio"),
-      feat("Lip Sync", "characteristic")
-    ],
-    paramConfig: {
-      ...params.imageInput(1, "Portrait Image", true),
-      ...params.resolution(["4k", "1080p", "720p"], "720p"),
-      ...params.aspectRatio(["16:9", "9:16", "4:5", "5:4", "1:1", "auto"]),
-      ...dynamicVoiceConfig,
-      ...params.prompt({ minLength: 20, maxLength: 5e3, placeholder: "Write the script your avatar will speak (at least 20 characters)..." })
-    }
-  },
-  // ── Video Avatar (t2v) ────────────────────────────────────────────
-  {
-    id: "heygen-video-avatar",
-    modelId: "heygen-avatar-iv",
-    addedAt: "2026-03-17",
-    name: "HeyGen Video Avatar",
-    workflow: "heygen/v1/video/generate",
-    buildPayload: buildHeyGenVideoAvatarPayload,
-    estimatedTime: 90,
-    mode: "video",
-    inputType: "t2v",
-    description: "Generate a speaking avatar video from a stock avatar and text script.",
-    features: [
-      feat("Avatar Selection", "characteristic"),
-      feat("Voice Selection", "audio"),
-      feat("Lip Sync", "characteristic")
-    ],
-    paramConfig: {
-      ...params.videoId([], "", {
-        required: true,
-        catalog: { workflow: "heygen/v1/catalog/avatars" }
-      }),
-      ...engineParam,
-      ...params.resolution(["4k", "1080p", "720p"], "720p"),
-      ...params.aspectRatio(["16:9", "9:16", "4:5", "5:4", "1:1", "auto"]),
-      ...dynamicVoiceConfig,
-      ...params.prompt({ minLength: 20, maxLength: 5e3, placeholder: "Write the script your avatar will speak (at least 20 characters)..." })
-    }
-  }
-]);
-
-// src/vendors/catalog/minimax.ts
-var buildMinimaxMusicPayload = (ctx) => ({
-  prompt: ctx.prompt,
-  ...ctx.lyricsPrompt ? { lyrics: ctx.lyricsPrompt } : {},
-  ...ctx.lyricsOptimizer != null ? { lyrics_optimizer: ctx.lyricsOptimizer } : {},
-  ...ctx.isInstrumental != null ? { is_instrumental: ctx.isInstrumental } : {},
-  audio_setting: {
-    sample_rate: ctx.sampleRate ?? 44100,
-    bitrate: ctx.bitrate ?? 256e3,
-    format: ctx.format ?? "mp3"
-  }
-});
-var H3_MAX_PROMPT_MAX = 5e4;
-var FRAME_REF_EXCLUSIVE2 = "References and start/end frames cannot be combined.";
-var AUDIO_NEEDS_VISUAL2 = "Audio cannot be the only reference \u2014 add an image or video.";
-var RATIO_FOLLOWS_FRAME = "Aspect ratio follows the frame image.";
-var ADAPTIVE_NEEDS_REFS = "Adaptive aspect ratio requires reference inputs.";
-var H3_MAX_RATIOS = ["21:9", "16:9", "4:3", "1:1", "3:4", "9:16"];
-var REFERENCE_EXCLUDES_FRAMES = {
-  startFrame: { disabled: true, reason: FRAME_REF_EXCLUSIVE2 },
-  endFrame: { disabled: true, reason: FRAME_REF_EXCLUSIVE2 }
-};
-var FRAME_EXCLUDES_REFERENCES = {
-  imageUrls: { disabled: true, reason: FRAME_REF_EXCLUSIVE2 },
-  videoUrls: { disabled: true, reason: FRAME_REF_EXCLUSIVE2 },
-  audioUrls: { disabled: true, reason: FRAME_REF_EXCLUSIVE2 }
-};
-var h3MaxConstraints = [
-  { when: { startFrame: { exists: true } }, then: FRAME_EXCLUDES_REFERENCES },
-  { when: { endFrame: { exists: true } }, then: FRAME_EXCLUDES_REFERENCES },
-  { when: { imageUrls: { exists: true } }, then: REFERENCE_EXCLUDES_FRAMES },
-  { when: { videoUrls: { exists: true } }, then: REFERENCE_EXCLUDES_FRAMES },
-  { when: { audioUrls: { exists: true } }, then: REFERENCE_EXCLUDES_FRAMES },
-  // With a frame the output follows that image, so the ratio is not a choice.
-  // An end frame alone is valid — it generates towards that keyframe.
-  { when: { startFrame: { exists: true } }, then: {
-    aspectRatio: { disabled: true, reason: RATIO_FOLLOWS_FRAME }
-  } },
-  { when: { endFrame: { exists: true } }, then: {
-    aspectRatio: { disabled: true, reason: RATIO_FOLLOWS_FRAME }
-  } },
-  // 'adaptive' means "follow the references" — no meaning without them.
-  { when: { imageUrls: { exists: false }, videoUrls: { exists: false }, audioUrls: { exists: false } }, then: {
-    aspectRatio: { allowed: [...H3_MAX_RATIOS], reason: ADAPTIVE_NEEDS_REFS }
-  } },
-  // Audio carries no visual signal, so it cannot be the only reference.
-  { when: { imageUrls: { exists: false }, videoUrls: { exists: false } }, then: {
-    audioUrls: { disabled: true, reason: AUDIO_NEEDS_VISUAL2 }
-  } }
-];
-var { MODELS: MODELS25 } = defineModels("minimax", [
-  // minimax-02-hd (minimax-tts) was removed in 6.0: its backend workflow was
-  // never deployed anywhere, so no generation ever existed to resolve — nothing
-  // to deprecate. Re-add as a fresh entry if MiniMax TTS ever ships.
-  {
-    id: "minimax-music-v2",
-    name: "MiniMax Music v2",
-    addedAt: "2026-02-06",
-    workflow: "minimax-music/v2",
-    buildPayload: buildMinimaxMusicPayload,
-    estimatedTime: 39,
-    mode: "audio",
-    inputType: "music",
-    description: "Text-to-music with vocals or instrumentals from a style prompt and lyrics prompt.",
-    features: [feat("Music", "characteristic"), feat("Vocals", "characteristic")],
-    paramConfig: {
-      ...params.prompt({ maxLength: 2e3, placeholder: "Describe the genre, mood, instruments, tempo, and production style..." }),
-      ...p.text("lyricsPrompt", {
-        maxLength: 3500,
-        label: "Lyrics",
-        placeholder: "Write lyrics, or describe the lyrical theme. Optional for instrumental or optimizer-generated lyrics."
-      }),
-      ...p.boolean("lyricsOptimizer", false, "Lyrics Optimizer"),
-      ...p.boolean("isInstrumental", false, "Instrumental"),
-      ...p.enum("sampleRate", [16e3, 24e3, 32e3, 44100], 44100, { label: "Sample Rate" }),
-      ...p.enum("bitrate", [32e3, 64e3, 128e3, 256e3], 256e3, { label: "Bitrate" }),
-      ...p.enum("format", ["mp3", "wav", "pcm"], "mp3", { label: "Format" })
-    }
-  },
-  {
-    id: "minimax-music-v3",
-    name: "MiniMax Music v3",
-    addedAt: "2026-08-14",
-    workflow: "minimax-music/v3",
-    estimatedTime: 40,
-    mode: "audio",
-    inputType: "music",
-    description: "Text-to-music with vocals or instrumentals from a style prompt and optional lyrics, with configurable audio encoding.",
-    features: [feat("Music", "characteristic"), feat("Vocals", "characteristic")],
-    paramConfig: {
-      ...params.prompt({ maxLength: 2e3, placeholder: "Describe the genre, mood, instruments, tempo, and production style..." }),
-      ...p.text("lyricsPrompt", {
-        maxLength: 3500,
-        label: "Lyrics",
-        placeholder: "Write lyrics; \\n separates lines, [Intro]/[Verse]/[Chorus] tags supported. Optional for instrumental or optimizer-generated lyrics."
-      }),
-      ...p.boolean("lyricsOptimizer", false, "Lyrics Optimizer"),
-      ...p.boolean("isInstrumental", false, "Instrumental"),
-      ...p.enum("sampleRate", [16e3, 24e3, 32e3, 44100], 44100, { label: "Sample Rate" }),
-      ...p.enum("bitrate", [32e3, 64e3, 128e3, 256e3], 256e3, { label: "Bitrate" }),
-      ...p.enum("format", ["mp3", "wav", "pcm"], "mp3", { label: "Format" })
-    }
-  },
-  {
-    // One entry for all three H3 Max modes: the inputs pick it — reference
-    // images, videos or audio generate from references; a start and/or end
-    // frame animates those frames; neither generates from the prompt alone.
-    id: "minimax-h3-max",
-    name: "MiniMax H3 Max",
-    modelId: "fal-ai-h3-max",
-    addedAt: "2026-08-28",
-    workflow: "minimax/h3-max/video-generation",
-    estimatedTime: 5,
-    mode: "video",
-    inputType: "t2v",
-    description: "Top-tier MiniMax H3 Max video from text, a start/end frame, or reference images, videos, and audio \u2014 refer to references in the prompt as Image 1, Video 1, Audio 1, in input order. Up to 15s at 1080p.",
-    features: [
-      feat("Start/End Frame", "frame"),
-      feat("Multi-Image Input", "input"),
-      feat("Video Input", "input"),
-      feat("Audio Input", "input"),
-      feat("1080p", "resolution"),
-      feat("5-15 sec", "duration")
-    ],
-    paramConfig: {
-      ...params.prompt({ maxLength: H3_MAX_PROMPT_MAX }),
-      ...params.startFrame(),
-      ...params.endFrame(),
-      // Reference slots — reference-to-video route. Clips are 2-15s each
-      // (≤15s combined per modality) and images + videos + audios must add up
-      // to ≤12 files — backend-enforced; only the per-array maxima live here.
-      ...params.imageInput(9, "Reference Images"),
-      ...params.videoInputs(3, "Reference Videos"),
-      ...params.audioInputs(3, "Reference Audios"),
-      // 1080p is a latent refinement of a native 768p generation.
-      ...params.resolution(["480p", "768p", "1080p"], "768p"),
-      ...params.durationRange(5, 15, 5),
-      // 'adaptive' means "follow the references" — constrained below.
-      ...params.aspectRatio(["adaptive", "21:9", "16:9", "4:3", "1:1", "3:4", "9:16"], "16:9"),
-      ...p.enum("promptExpansionMode", ["balanced", "quality"], "balanced", { label: "Prompt Expansion" }),
-      // -1 (sentinel) means "pick a random seed"; the builder drops it.
-      ...p.range("seed", -1, 2147483647, -1),
-      ...p.boolean("enableSafetyChecker", true, "Safety Checker")
-    },
-    constraints: h3MaxConstraints
-  },
-  {
-    // Turbo sibling of minimax-h3-max — same model family, tuned for speed
-    // (faster-than-realtime generation), text and frames only.
-    id: "minimax-h3-max-turbo",
-    name: "MiniMax H3 Max Turbo",
-    modelId: "fal-ai-h3-max-turbo",
-    addedAt: "2026-09-03",
-    workflow: "minimax/h3-max-turbo/text-to-video",
-    editWorkflow: "minimax/h3-max-turbo/image-to-video",
-    estimatedTime: 5,
-    mode: "video",
-    inputType: "t2v",
-    description: "Faster-than-realtime MiniMax H3 Max Turbo video from text or a start/end frame, with prompt expansion. Up to 15s at 1080p.",
-    features: [
-      feat("Fast", "characteristic"),
-      feat("Image Input", "input"),
-      feat("Start/End Frame", "frame"),
-      feat("1080p", "resolution"),
-      feat("5-15 sec", "duration")
-    ],
-    paramConfig: {
-      ...params.prompt({ maxLength: H3_MAX_PROMPT_MAX }),
-      ...params.startFrame(),
-      ...params.endFrame(),
-      ...params.resolution(["480p", "768p", "1080p"], "768p"),
-      ...params.durationRange(5, 15, 5),
-      ...params.aspectRatio(["21:9", "16:9", "4:3", "1:1", "3:4", "9:16"], "16:9"),
-      ...p.enum("promptExpansionMode", ["balanced", "quality"], "balanced", { label: "Prompt Expansion" }),
-      // -1 (sentinel) means "pick a random seed"; the builder drops it.
-      ...p.range("seed", -1, 2147483647, -1),
-      ...p.boolean("enableSafetyChecker", true, "Safety Checker")
-    },
-    constraints: [
-      // Either frame routes to image-to-video, which derives the ratio from
-      // the image it was given (no aspect_ratio on that wire). An end frame
-      // alone is valid — fal documents it as end-only keyframe generation.
-      { when: { startFrame: { exists: true } }, then: {
-        aspectRatio: { disabled: true, reason: "Aspect ratio follows the frame image." }
-      } },
-      { when: { endFrame: { exists: true } }, then: {
-        aspectRatio: { disabled: true, reason: "Aspect ratio follows the frame image." }
-      } }
-    ]
+      },
+      ...params.prompt({ minLength: 5, maxLength: 1e3 })
+    }
   },
   {
-    // Camera-controls sibling of minimax-h3-max: animates a single start
-    // frame along scripted camera keyframes. The prompt is optional — blank
-    // freezes the scene and moves only the camera. No aspect ratio on this
-    // wire; the output follows the frame image.
-    id: "minimax-h3-max-camera-controls",
-    name: "MiniMax H3 Max Camera Controls",
-    modelId: "fal-ai-h3-max-camera-controls",
-    addedAt: "2026-09-14",
-    workflow: "minimax/h3-max/camera-controls",
-    estimatedTime: 5,
+    id: "eleven-voice-design-v3",
+    name: "Eleven Voice Design v3",
+    addedAt: "2026-03-24",
+    modelId: "eleven_ttv_v3",
+    workflow: "elevenlabs/v1/voice-design",
+    buildPayload: buildElevenLabsVoiceDesignPayload("eleven_ttv_v3"),
+    estimatedTime: 15,
+    mode: "audio",
+    inputType: "tts",
+    description: "Design a new voice from a text description using v3 engine.",
+    features: [feat("Voice Design", "characteristic"), feat("Preview", "characteristic")],
+    paramConfig: { ...params.prompt({ minLength: 20, maxLength: 1e3 }) }
+  },
+  {
+    id: "eleven-voice-design-v2",
+    name: "Eleven Voice Design Multilingual v2",
+    addedAt: "2026-03-24",
+    modelId: "eleven_multilingual_ttv_v2",
+    workflow: "elevenlabs/v1/voice-design",
+    buildPayload: buildElevenLabsVoiceDesignPayload("eleven_multilingual_ttv_v2"),
+    estimatedTime: 15,
+    mode: "audio",
+    inputType: "tts",
+    description: "Design a new voice from a text description with multilingual support.",
+    features: [feat("Voice Design", "characteristic"), feat("Multilingual", "characteristic"), feat("Preview", "characteristic")],
+    paramConfig: { ...params.prompt({ minLength: 20, maxLength: 1e3 }) }
+  },
+  {
+    id: "eleven-voice-create",
+    name: "Eleven Voice Previews",
+    // Pricing registers this operation as `eleven-voice-create-previews`, which
+    // does not match the model id — without the hint the credits lookup misses.
+    modelId: "eleven-voice-create-previews",
+    addedAt: "2026-03-24",
+    workflow: "elevenlabs/v1/voice-create-previews",
+    buildPayload: buildElevenLabsVoicePreviewsPayload,
+    estimatedTime: 15,
+    mode: "audio",
+    inputType: "tts",
+    description: "Generate voice previews from a description to audition before committing.",
+    features: [feat("Voice Design", "characteristic"), feat("Preview", "characteristic")],
+    paramConfig: { ...params.prompt({ minLength: 20, maxLength: 1e3 }) }
+  }
+]);
+
+// src/vendors/catalog/elevenlabs.payloads.ts
+var buildElevenLabsMusicPayload = (input) => ({
+  prompt: input.prompt,
+  music_length_seconds: input.duration ?? 30,
+  model_id: "music_v2",
+  force_instrumental: input.isInstrumental ?? false
+});
+var voiceSettings = (input) => {
+  const settings = {
+    ...input.stability != null ? { stability: input.stability } : {},
+    ...input.similarityBoost != null ? { similarity_boost: input.similarityBoost } : {},
+    ...input.styleExaggeration != null ? { style: input.styleExaggeration } : {},
+    ...input.speed != null ? { speed: input.speed } : {},
+    ...input.useSpeakerBoost != null ? { use_speaker_boost: input.useSpeakerBoost } : {}
+  };
+  return Object.keys(settings).length > 0 ? { voice_settings: settings } : {};
+};
+var buildElevenLabsTTSPayload = (modelId) => (input) => ({
+  text: input.prompt,
+  voice_id: input.voiceId ?? DEFAULT_VOICE_ID,
+  model_id: modelId,
+  ...input.language ? { language_code: input.language } : {},
+  ...voiceSettings(input)
+});
+var buildElevenLabsSTSPayload = (modelId) => (input) => ({
+  audio_url: input.audioUrl,
+  voice_id: input.voiceId ?? DEFAULT_VOICE_ID,
+  model_id: modelId,
+  remove_background_noise: input.removeBackgroundNoise ?? false,
+  ...voiceSettings(input)
+});
+var buildElevenLabsDialoguePayload = (input) => ({
+  conversation: input.dialogue.map((line) => ({
+    voice_id: line.voiceId,
+    text: line.text
+  })),
+  model_id: "eleven_v3",
+  ...input.stability != null ? { settings: { stability: input.stability } } : {},
+  ...input.language ? { language_code: input.language } : {},
+  ...input.seed != null ? { seed: input.seed } : {}
+});
+registerPayloads(MODELS24, {
+  "elevenlabs-music-v2": buildElevenLabsMusicPayload,
+  "eleven-v3": buildElevenLabsTTSPayload("eleven_v3"),
+  "eleven-multilingual-v2": buildElevenLabsTTSPayload("eleven_multilingual_v2"),
+  "eleven-sts-v2": buildElevenLabsSTSPayload("eleven_english_sts_v2"),
+  "eleven-multilingual-sts-v2": buildElevenLabsSTSPayload("eleven_multilingual_sts_v2"),
+  "eleven-text-to-dialogue": buildElevenLabsDialoguePayload
+});
+
+// src/vendors/catalog/heygen.ts
+var buildHeyGenPhotoAvatarPayload = (ctx) => ({
+  image_url: ctx.imageUrls?.[0],
+  script: ctx.prompt,
+  voice_id: ctx.voiceId || void 0,
+  ...ctx.resolution ? { resolution: ctx.resolution } : {},
+  ...ctx.aspectRatio ? { aspect_ratio: ctx.aspectRatio } : {}
+});
+var buildHeyGenVideoAvatarPayload = (ctx) => {
+  const c = ctx;
+  return {
+    avatar_id: ctx.videoId || void 0,
+    script: ctx.prompt,
+    voice_id: ctx.voiceId || void 0,
+    ...ctx.resolution ? { resolution: ctx.resolution } : {},
+    ...ctx.aspectRatio ? { aspect_ratio: ctx.aspectRatio } : {},
+    ...c.engine ? { engine: c.engine } : {}
+  };
+};
+var engineParam = {
+  engine: {
+    label: "Engine",
+    required: false,
+    descriptor: {
+      kind: "enum",
+      valueType: "string",
+      options: [
+        { id: "avatar_iv", label: "Avatar IV" },
+        { id: "avatar_v", label: "Avatar V" }
+      ],
+      default: "avatar_iv"
+    }
+  }
+};
+var dynamicVoiceConfig = {
+  ...params.voiceId([], "", {
+    required: true,
+    catalog: { workflow: "heygen/v1/catalog/voices" }
+  })
+};
+var { MODELS: MODELS25 } = defineModels("heygen", [
+  // ── Photo Avatar (i2v) ────────────────────────────────────────────
+  {
+    id: "heygen-talking-photo",
+    modelId: "heygen-avatar-iv",
+    addedAt: "2026-03-17",
+    name: "HeyGen Talking Photo",
+    workflow: "heygen/v1/video/generate",
+    buildPayload: buildHeyGenPhotoAvatarPayload,
+    estimatedTime: 120,
     mode: "video",
     inputType: "i2v",
-    description: "MiniMax H3 Max video from a start frame with scripted camera motion \u2014 2-12 keyframes set the camera angle, height, and distance over time. Leave the prompt blank to freeze the scene and move only the camera. Up to 15s at 1080p.",
+    description: "Animate any photo into a speaking avatar with natural lip-sync.",
     features: [
-      feat("Camera Controls", "characteristic"),
       feat("Image Input", "input"),
-      feat("1080p", "resolution"),
-      feat("5-15 sec", "duration")
+      feat("Voice Selection", "audio"),
+      feat("Lip Sync", "characteristic")
     ],
     paramConfig: {
-      ...params.prompt({
-        maxLength: H3_MAX_PROMPT_MAX,
-        required: false,
-        placeholder: "Optional \u2014 leave blank to freeze the scene and move only the camera..."
-      }),
-      ...params.startFrame("Start Frame", true),
-      // Native 480p default here (the siblings default to 768p); 1080p is a
-      // latent refinement of a native 768p generation.
-      ...params.resolution(["480p", "768p", "1080p"], "480p"),
-      ...params.durationRange(5, 15, 5),
-      // Ordered camera keyframes: the first pose is held before its time and
-      // the final pose for the remainder. Azimuth keeps signed full turns,
-      // capped at ±32 turns of total travel; distance is in normalized scene
-      // units and must stay above zero (the wire has no upper bound).
-      cameraTrajectory: {
-        label: "Camera Trajectory",
-        descriptor: {
-          kind: "object",
-          array: { min: 2, max: 12 },
-          fields: {
-            time: { kind: "range", min: 0, max: 1, step: 0.01, default: 0 },
-            azimuth: { kind: "range", min: -11520, max: 11520, default: 0 },
-            elevation: { kind: "range", min: -90, max: 90, default: 0 },
-            distance: { kind: "range", min: 0.01, max: 100, default: 1 }
-          }
-        }
-      },
-      ...p.enum("promptExpansionMode", ["balanced", "quality"], "balanced", { label: "Prompt Expansion" }),
-      // -1 (sentinel) means "pick a random seed"; the builder drops it.
-      ...p.range("seed", -1, 2147483647, -1),
-      ...p.boolean("enableSafetyChecker", true, "Safety Checker")
+      ...params.imageInput(1, "Portrait Image", true),
+      ...params.resolution(["4k", "1080p", "720p"], "720p"),
+      ...params.aspectRatio(["16:9", "9:16", "4:5", "5:4", "1:1", "auto"]),
+      ...dynamicVoiceConfig,
+      ...params.prompt({ minLength: 20, maxLength: 5e3, placeholder: "Write the script your avatar will speak (at least 20 characters)..." })
     }
   },
+  // ── Video Avatar (t2v) ────────────────────────────────────────────
   {
-    // Lip-sync sibling of minimax-h3-max: animates a portrait to a supplied
-    // soundtrack. No prompt and no duration on this wire — the audio drives
-    // both: the vendor requires at least 5s, clips anything past 14.8s, and
-    // the output matches the clipped track. The output follows the image, so
-    // there is no aspect ratio either.
-    id: "minimax-h3-max-lip-sync",
-    name: "MiniMax H3 Max Lip Sync",
-    modelId: "fal-ai-h3-max-lip-sync",
-    addedAt: "2026-09-22",
-    workflow: "minimax/h3-max/lip-sync/image-to-video",
-    estimatedTime: 120,
+    id: "heygen-video-avatar",
+    modelId: "heygen-avatar-iv",
+    addedAt: "2026-03-17",
+    name: "HeyGen Video Avatar",
+    workflow: "heygen/v1/video/generate",
+    buildPayload: buildHeyGenVideoAvatarPayload,
+    estimatedTime: 90,
     mode: "video",
-    inputType: "i2v",
-    description: "MiniMax H3 Max lip-synced video from a portrait and an audio track \u2014 mouth movements follow the soundtrack, optionally guided by a transcript. 5-14.8s of audio, up to 2K.",
+    inputType: "t2v",
+    description: "Generate a speaking avatar video from a stock avatar and text script.",
     features: [
-      feat("Lip Sync", "characteristic"),
-      feat("Image Input", "input"),
-      feat("Audio Input", "audio"),
-      feat("2K", "resolution"),
-      feat("5-15 sec", "duration")
+      feat("Avatar Selection", "characteristic"),
+      feat("Voice Selection", "audio"),
+      feat("Lip Sync", "characteristic")
     ],
     paramConfig: {
-      ...params.startFrame("Portrait Image", true),
-      ...params.audioInput("Audio Track", true),
-      // Unlike the siblings this one also offers 2K; 1080p and 2K are latent
-      // refinements of a native 768p generation.
-      ...params.resolution(["480p", "768p", "1080p", "2k"], "768p"),
-      // Transcribes the audio to guide the sync; off follows the waveform alone.
-      ...p.boolean("enableTranscription", false, "Transcription"),
-      // -1 (sentinel) means "pick a random seed"; the builder drops it.
-      ...p.range("seed", -1, 2147483647, -1),
-      ...p.boolean("enableSafetyChecker", true, "Safety Checker")
+      ...params.videoId([], "", {
+        required: true,
+        catalog: { workflow: "heygen/v1/catalog/avatars" }
+      }),
+      ...engineParam,
+      ...params.resolution(["4k", "1080p", "720p"], "720p"),
+      ...params.aspectRatio(["16:9", "9:16", "4:5", "5:4", "1:1", "auto"]),
+      ...dynamicVoiceConfig,
+      ...params.prompt({ minLength: 20, maxLength: 5e3, placeholder: "Write the script your avatar will speak (at least 20 characters)..." })
     }
   }
 ]);
@@ -7417,14 +7432,14 @@ var buildMinimaxH3MaxLipSyncPayload = (input) => ({
   ...input.seed != null && input.seed !== -1 ? { seed: input.seed } : {},
   enable_safety_checker: input.enableSafetyChecker ?? true
 });
-registerPayloads(MODELS25, {
+registerPayloads(MODELS9, {
   "minimax-music-v3": buildMinimaxMusicV3Payload,
   "minimax-h3-max": buildMinimaxH3MaxPayload,
   "minimax-h3-max-turbo": buildMinimaxH3MaxTurboPayload,
   "minimax-h3-max-camera-controls": buildMinimaxH3MaxCameraControlsPayload,
   "minimax-h3-max-lip-sync": buildMinimaxH3MaxLipSyncPayload
 });
-registerEditPayloads(MODELS25, {
+registerEditPayloads(MODELS9, {
   "minimax-h3-max-turbo": buildMinimaxH3MaxTurboPayload
 });
 
@@ -9994,7 +10009,6 @@ var ALL_MODELS = [
   ...MODELS6,
   ...MODELS7,
   ...MODELS8,
-  ...MODELS9,
   ...MODELS10,
   ...MODELS11,
   ...MODELS12,
@@ -10011,6 +10025,7 @@ var ALL_MODELS = [
   ...MODELS23,
   ...MODELS24,
   ...MODELS25,
+  ...MODELS9,
   ...MODELS26,
   ...MODELS31,
   ...MODELS27,
@@ -11573,6 +11588,8 @@ var MAX_DRIVE_PROMPT_LENGTH = 18e3;
 
 // src/client/drive.ts
 var USER_REACTION_ATTR = "userReaction";
+var DRAFT_TASK_ATTR = "draft_task";
+var DRAFT_EXPIRES_AT_ATTR = "draft_expires_at";
 function inferResourceType(mode) {
   if (mode === "video") return "VIDEO";
   if (mode === "audio") return "AUDIO";
@@ -11619,6 +11636,14 @@ function parseAttributes(raw) {
   }
   return map;
 }
+function parseDraftInfo(attrs) {
+  const draftTask = parseJsonAttr(attrs[DRAFT_TASK_ATTR]);
+  if (!isSeedanceDraftTask(draftTask)) return {};
+  return {
+    draftTask: { id: draftTask.id, video_input: draftTask.video_input, signature: draftTask.signature },
+    ...attrs[DRAFT_EXPIRES_AT_ATTR] ? { draftExpiresAt: attrs[DRAFT_EXPIRES_AT_ATTR] } : {}
+  };
+}
 function parseReaction(value) {
   return value === "like" || value === "dislike" ? value : void 0;
 }
@@ -11691,7 +11716,8 @@ function toDetailedItem(file) {
     referenceAudioUrl: extras.referenceAudioUrl,
     aspectRatio: extras.aspectRatio,
     resolution: extras.resolution,
-    quality: extras.quality
+    quality: extras.quality,
+    ...parseDraftInfo(attrs)
   };
 }
 var LEGACY_TOOL_APP = {
@@ -11736,7 +11762,8 @@ function adaptLegacyGeneration(attrs) {
     appType: app?.appType,
     model: attrs.model || void 0,
     aiSDKPayload,
-    userReaction: parseReaction(attrs[USER_REACTION_ATTR])
+    userReaction: parseReaction(attrs[USER_REACTION_ATTR]),
+    ...parseDraftInfo(attrs)
   };
 }
 function parseGeneration(file) {
@@ -11749,7 +11776,8 @@ function parseGeneration(file) {
     appType: attrs.appType === "native" || attrs.appType === "miniapp" ? attrs.appType : void 0,
     model: attrs.model || void 0,
     aiSDKPayload: parseJsonAttr(attrs.aiSDKPayload),
-    userReaction: parseReaction(attrs[USER_REACTION_ATTR])
+    userReaction: parseReaction(attrs[USER_REACTION_ATTR]),
+    ...parseDraftInfo(attrs)
   };
 }
 function createDriveClient(f, apiUrl, rootFolderName) {