@picsart/ai-sdk 5.22.1 → 5.24.0

package/README.md

@@ -144,10 +144,10 @@ catalog.find({ output: 'text' })
 image/video model throws, and `generate()` throws on a text model — use the matching
 method for each.
 
-## Voice & Avatar Catalogs
+## Voice, Avatar & Template Catalogs
 
-Models with catalog-backed params (voices, avatars) serve their option lists
-from platform catalog tasks (`<vendor>/v1/catalog/<voices|avatars>`) — nothing
+Models with catalog-backed params (voices, avatars, effect / caption templates) serve their option lists
+from platform catalog tasks (`<vendor>/v1/catalog/<voices|avatars|templates|…>`) — nothing
 is bundled; the workers cache the lists and answer fast. Fetch them via
 `ai.catalogs`:
 

package/index.d.ts

@@ -112,6 +112,10 @@ type ModelInputById = {
     "bytedance-video-upscaler": {
         videoUrl: string;
     };
+    "captionsai-video-captions": {
+        videoUrl: string;
+        templateId?: string;
+    };
     "claude-haiku-4-5": {
         prompt: string;
         imageUrls?: string[];
@@ -1824,6 +1828,15 @@ interface FileDescriptor {
      * backend worker stays the authoritative gate. Omit for no client-side floor.
      */
     minPixels?: number;
+    /**
+     * Min intrinsic short-side length (pixels) accepted for an image/video file —
+     * `min(width, height)` must be at least this. Most vendors specify input limits
+     * per side rather than by total pixel count (e.g. Seedance reference images:
+     * width and height each in [300, 6000]). Enforced client-side at upload by
+     * measuring the media before it is sent; the backend worker stays the
+     * authoritative gate. Omit for no client-side floor.
+     */
+    minSidePixels?: number;
     /**
      * Max intrinsic short-side length (pixels) accepted for an image/video file —
      * `min(width, height)` must not exceed this. Used by upscalers whose source
@@ -2013,7 +2026,7 @@ interface ModelFilter$1 {
     release?: ReleaseTag[];
 }
 
-type AppProvider = 'picsart' | 'google' | 'kling' | 'grok' | 'openai' | 'flux' | 'ideogram' | 'elevenlabs' | 'minimax' | 'wan' | 'seedance' | 'ltx' | 'seedream' | 'seedaudio' | 'hunyuan' | 'pika' | 'runway' | 'luma' | 'ovi' | 'creatify' | 'veed' | 'bytedance' | 'qwen' | 'reve' | 'recraft' | 'videography' | 'topaz' | 'heygen' | 'happyhorse' | 'pixverse' | 'anthropic' | 'async';
+type AppProvider = 'picsart' | 'google' | 'kling' | 'grok' | 'openai' | 'flux' | 'ideogram' | 'elevenlabs' | 'minimax' | 'wan' | 'seedance' | 'ltx' | 'seedream' | 'seedaudio' | 'hunyuan' | 'pika' | 'runway' | 'luma' | 'ovi' | 'creatify' | 'veed' | 'bytedance' | 'qwen' | 'reve' | 'recraft' | 'videography' | 'topaz' | 'heygen' | 'happyhorse' | 'pixverse' | 'anthropic' | 'async' | 'captionsai';
 /** Provider used by model definitions. */
 type Provider = AppProvider;
 /** App generation modes. */
@@ -2514,6 +2527,7 @@ declare const Models: {
     readonly BytedanceOmnihumanV15: "bytedance-omnihuman-v1.5";
     readonly BytedanceVideoEnhance: "bytedance-video-enhance";
     readonly BytedanceVideoUpscaler: "bytedance-video-upscaler";
+    readonly CaptionsaiVideoCaptions: "captionsai-video-captions";
     readonly ClaudeHaiku45: "claude-haiku-4-5";
     readonly ClaudeOpus48: "claude-opus-4-8";
     readonly ClaudeSonnet46: "claude-sonnet-4-6";

package/index.js

@@ -820,7 +820,8 @@ var providers = {
   happyhorse: { color: "#FF6A00", label: "HH", name: "Happy Horse" },
   pixverse: { color: "#7C3AED", label: "PV", name: "PixVerse" },
   anthropic: { color: "#D97757", label: "CL", name: "Anthropic" },
-  async: { color: "#5E5CE6", label: "AA", name: "Async AI" }
+  async: { color: "#5E5CE6", label: "AA", name: "Async AI" },
+  captionsai: { color: "#1D1F20", label: "MR", name: "Mirage" }
 };
 
 // src/core/descriptors/presets.ts
@@ -1018,6 +1019,7 @@ var p = {
           ...opts?.array ? { array: opts.array } : {},
           ...opts?.maxDurationSec != null ? { maxDurationSec: opts.maxDurationSec } : {},
           ...opts?.minPixels != null ? { minPixels: opts.minPixels } : {},
+          ...opts?.minSidePixels != null ? { minSidePixels: opts.minSidePixels } : {},
           ...opts?.maxShortSidePixels != null ? { maxShortSidePixels: opts.maxShortSidePixels } : {},
           ...opts?.maxBytes != null ? { maxBytes: opts.maxBytes } : {}
         }
@@ -1175,7 +1177,7 @@ var passthroughPayload = (paramConfig) => (ctx) => {
   return payload;
 };
 function defineModels(provider, configs) {
-  const MODELS37 = [];
+  const MODELS38 = [];
   for (const c of configs) {
     const prov = c.provider ?? provider;
     const resolvedPayload = c.buildPayload ?? passthroughPayload(c.paramConfig);
@@ -1211,19 +1213,19 @@ function defineModels(provider, configs) {
     if (c.constraints !== void 0) model.constraints = c.constraints;
     const contract = createModelContract(model);
     model.outputSchema = c.outputSchema ?? contract.output;
-    MODELS37.push(model);
+    MODELS38.push(model);
   }
-  return { MODELS: MODELS37 };
+  return { MODELS: MODELS38 };
 }
-function registerPayloads(MODELS37, payloads) {
+function registerPayloads(MODELS38, payloads) {
   for (const [id, builder] of Object.entries(payloads)) {
-    const model = MODELS37.find((m) => m.id === id);
+    const model = MODELS38.find((m) => m.id === id);
     if (model) model.buildPayload = builder;
   }
 }
-function registerEditPayloads(MODELS37, payloads) {
+function registerEditPayloads(MODELS38, payloads) {
   for (const [id, builder] of Object.entries(payloads)) {
-    const model = MODELS37.find((m) => m.id === id);
+    const model = MODELS38.find((m) => m.id === id);
     if (model) model.buildEditPayload = builder;
   }
 }
@@ -1251,7 +1253,18 @@ var params = {
   // `category` defaults to the most common role for the slot (overridable per call):
   //   asset    → start/end frame, sync audio (direct inputs to the output)
   //   reference → ref images/videos/audios (guidance signals)
-  imageInput: (max = 1, label = "Start Image", required = false, category = "reference", minPixels) => p.file("imageUrls", "image", { array: { max }, label, required, category, ...minPixels != null ? { minPixels } : {} }),
+  /** Array of image inputs (writes to `imageUrls`). `bounds` carries the
+   *  client-side dimension floors enforced at upload: `minPixels` for a vendor
+   *  rule stated as a total pixel count, `minSidePixels` for one stated per
+   *  side (width and height each), which is what most vendors publish. */
+  imageInput: (max = 1, label = "Start Image", required = false, category = "reference", bounds) => p.file("imageUrls", "image", {
+    array: { max },
+    label,
+    required,
+    category,
+    ...bounds?.minPixels != null ? { minPixels: bounds.minPixels } : {},
+    ...bounds?.minSidePixels != null ? { minSidePixels: bounds.minSidePixels } : {}
+  }),
   /** Single source-video slot (v2v / video edit). Writes to `videoUrl`.
    *  `maxDurationSec` caps the source clip length, `maxShortSidePixels` caps
    *  the shorter side (upscaler sources) and `maxBytes` caps the file size,
@@ -1267,15 +1280,17 @@ var params = {
   /** Single driving / sync-audio slot. Writes to `audioUrl`. */
   audioInput: (label = "Audio Track", required = false, category = "asset") => p.file("audioUrl", "audio", { label, required, category }),
   /** Array of reference videos (writes to `videoUrls`). Backend enforces
-   *  per-model total-duration caps (e.g. ≤ 15s for seedance). `maxBytes` caps
-   *  each individual clip's file size, enforced client-side at upload. */
-  videoInputs: (max = 3, label = "Reference Videos", required = false, minPixels, maxBytes) => p.file("videoUrls", "video", {
+   *  per-model total-duration caps (e.g. ≤ 15s for seedance). `bounds` carries
+   *  the client-side checks run at upload: `minPixels` for a total-pixel floor,
+   *  `minSidePixels` for a per-side one, `maxBytes` for each clip's file size. */
+  videoInputs: (max = 3, label = "Reference Videos", required = false, bounds) => p.file("videoUrls", "video", {
     array: { max },
     label,
     required,
     category: "reference",
-    ...minPixels != null ? { minPixels } : {},
-    ...maxBytes != null ? { maxBytes } : {}
+    ...bounds?.minPixels != null ? { minPixels: bounds.minPixels } : {},
+    ...bounds?.minSidePixels != null ? { minSidePixels: bounds.minSidePixels } : {},
+    ...bounds?.maxBytes != null ? { maxBytes: bounds.maxBytes } : {}
   }),
   /** Array of reference audios (writes to `audioUrls`). Backend enforces
    *  per-model total-duration caps. */
@@ -3540,7 +3555,8 @@ registerPayloads(MODELS11, {
 
 // src/vendors/catalog/seedance.ts
 var SEEDANCE_FRAME_REF_REASON = "Start/End frames cannot be combined with reference images, videos, or audios";
-var SEEDANCE_MIN_PIXELS = 409600;
+var SEEDANCE_MIN_SIDE_PIXELS = 300;
+var SEEDANCE_VIDEO_MIN_PIXELS = 407696;
 var SEEDANCE_25_MAX_VIDEO_BYTES = 209715200;
 var seedance20Constraints = [
   {
@@ -3791,8 +3807,12 @@ var { MODELS: MODELS12 } = defineModels("seedance", [
       ...params.returnLastFrame(),
       ...p.enum("outputFormat", ["mp4", "mov"], "mp4", { label: "Format" }),
       // 2.5 lifts the reference caps to 30 images / 10 videos / 10 audios.
-      ...params.imageInput(30, "Reference Images", false, "reference", SEEDANCE_MIN_PIXELS),
-      ...params.videoInputs(10, "Reference Videos", false, SEEDANCE_MIN_PIXELS, SEEDANCE_25_MAX_VIDEO_BYTES),
+      ...params.imageInput(30, "Reference Images", false, "reference", { minSidePixels: SEEDANCE_MIN_SIDE_PIXELS }),
+      ...params.videoInputs(10, "Reference Videos", false, {
+        minPixels: SEEDANCE_VIDEO_MIN_PIXELS,
+        minSidePixels: SEEDANCE_MIN_SIDE_PIXELS,
+        maxBytes: SEEDANCE_25_MAX_VIDEO_BYTES
+      }),
       ...params.audioInputs(10, "Reference Audios"),
       ...params.startFrame(),
       ...params.endFrame()
@@ -3846,7 +3866,7 @@ var { MODELS: MODELS12 } = defineModels("seedance", [
       ...params.durationRange(SEEDANCE_25_DURATION.min, SEEDANCE_25_DURATION.max, 15),
       ...params.generateAudio(),
       ...p.enum("outputFormat", ["mp4", "mov"], "mp4", { label: "Format" }),
-      ...params.videoInputs(10, "Source Videos", true, void 0, SEEDANCE_25_MAX_VIDEO_BYTES)
+      ...params.videoInputs(10, "Source Videos", true, { maxBytes: SEEDANCE_25_MAX_VIDEO_BYTES })
     }
   },
   {
@@ -3872,8 +3892,11 @@ var { MODELS: MODELS12 } = defineModels("seedance", [
       ...params.returnLastFrame(),
       // Reference roles map directly to backend `reference_*` content entries.
       // start/end frame stay on their own named slots.
-      ...params.imageInput(9, "Reference Images", false, "reference", SEEDANCE_MIN_PIXELS),
-      ...params.videoInputs(3, "Reference Videos", false, SEEDANCE_MIN_PIXELS),
+      ...params.imageInput(9, "Reference Images", false, "reference", { minSidePixels: SEEDANCE_MIN_SIDE_PIXELS }),
+      ...params.videoInputs(3, "Reference Videos", false, {
+        minPixels: SEEDANCE_VIDEO_MIN_PIXELS,
+        minSidePixels: SEEDANCE_MIN_SIDE_PIXELS
+      }),
       ...params.audioInputs(3, "Reference Audios"),
       ...params.startFrame(),
       ...params.endFrame()
@@ -3906,8 +3929,11 @@ var { MODELS: MODELS12 } = defineModels("seedance", [
       ...params.returnLastFrame(),
       // Reference roles map directly to backend `reference_*` content entries.
       // start/end frame stay on their own named slots.
-      ...params.imageInput(9, "Reference Images", false, "reference", SEEDANCE_MIN_PIXELS),
-      ...params.videoInputs(3, "Reference Videos", false, SEEDANCE_MIN_PIXELS),
+      ...params.imageInput(9, "Reference Images", false, "reference", { minSidePixels: SEEDANCE_MIN_SIDE_PIXELS }),
+      ...params.videoInputs(3, "Reference Videos", false, {
+        minPixels: SEEDANCE_VIDEO_MIN_PIXELS,
+        minSidePixels: SEEDANCE_MIN_SIDE_PIXELS
+      }),
       ...params.audioInputs(3, "Reference Audios"),
       ...params.startFrame(),
       ...params.endFrame()
@@ -3936,8 +3962,11 @@ var { MODELS: MODELS12 } = defineModels("seedance", [
       ...params.returnLastFrame(),
       // Reference roles map directly to backend `reference_*` content entries.
       // start/end frame stay on their own named slots.
-      ...params.imageInput(9, "Reference Images", false, "reference", SEEDANCE_MIN_PIXELS),
-      ...params.videoInputs(3, "Reference Videos", false, SEEDANCE_MIN_PIXELS),
+      ...params.imageInput(9, "Reference Images", false, "reference", { minSidePixels: SEEDANCE_MIN_SIDE_PIXELS }),
+      ...params.videoInputs(3, "Reference Videos", false, {
+        minPixels: SEEDANCE_VIDEO_MIN_PIXELS,
+        minSidePixels: SEEDANCE_MIN_SIDE_PIXELS
+      }),
       ...params.audioInputs(3, "Reference Audios"),
       ...params.startFrame(),
       ...params.endFrame()
@@ -3966,8 +3995,11 @@ var { MODELS: MODELS12 } = defineModels("seedance", [
       ...params.returnLastFrame(),
       // Reference roles map directly to backend `reference_*` content entries.
       // start/end frame stay on their own named slots.
-      ...params.imageInput(9, "Reference Images", false, "reference", SEEDANCE_MIN_PIXELS),
-      ...params.videoInputs(3, "Reference Videos", false, SEEDANCE_MIN_PIXELS),
+      ...params.imageInput(9, "Reference Images", false, "reference", { minSidePixels: SEEDANCE_MIN_SIDE_PIXELS }),
+      ...params.videoInputs(3, "Reference Videos", false, {
+        minPixels: SEEDANCE_VIDEO_MIN_PIXELS,
+        minSidePixels: SEEDANCE_MIN_SIDE_PIXELS
+      }),
       ...params.audioInputs(3, "Reference Audios"),
       ...params.startFrame(),
       ...params.endFrame()
@@ -8481,6 +8513,54 @@ registerPayloads(MODELS36, {
   "gemini-3.5-flash-lite": buildOpenAiPayload("gemini-3.5-flash-lite")
 });
 
+// src/vendors/catalog/captionsai.ts
+var CAPTIONS_MAX_DURATION_SEC = 300;
+var CAPTIONS_MAX_BYTES = 50 * 1024 * 1024;
+var DEFAULT_CAPTION_TEMPLATE_ID = "ctpl_DxflLOnuKkb198FNdI9E";
+var { MODELS: MODELS37 } = defineModels("captionsai", [
+  {
+    id: "captionsai-video-captions",
+    name: "Captions",
+    modelId: "mirage-captions",
+    addedAt: "2026-08-27",
+    workflow: "captionsai/v1/videos/captions",
+    // ~26s measured for an 8s clip on the live API; scales with clip length.
+    estimatedTime: 60,
+    mode: "video",
+    inputType: "v2v",
+    // Stage-only until the worker is deployed to prod and the pricing record
+    // (mirage-captions / video-to-video) exists — flip to production then.
+    release: "preview",
+    description: "Auto-transcribes a vertical video and burns in animated captions from 67 style templates \u2014 up to 5 minutes, 9:16.",
+    features: [
+      feat("Video Required", "input"),
+      feat("9:16", "resolution"),
+      feat("Up to 5 min", "duration"),
+      feat("67 Templates", "style")
+    ],
+    paramConfig: {
+      ...params.videoInput("Source Video", "asset", true, CAPTIONS_MAX_DURATION_SEC, void 0, CAPTIONS_MAX_BYTES),
+      ...params.catalog("templateId", {
+        label: "Caption Style",
+        // Not `required`: the declared default fills it, and request validation runs
+        // before the payload builder — a required flag would reject the very
+        // calls the default exists for (same shape as Kling's effect templateId).
+        source: { workflow: "captionsai/v1/catalog/caption-templates" },
+        default: DEFAULT_CAPTION_TEMPLATE_ID
+      })
+    }
+  }
+]);
+
+// src/vendors/catalog/captionsai.payloads.ts
+var buildCaptionsPayload = (input) => ({
+  video: { url: input.videoUrl },
+  caption_template_id: input.templateId ?? DEFAULT_CAPTION_TEMPLATE_ID
+});
+registerPayloads(MODELS37, {
+  "captionsai-video-captions": buildCaptionsPayload
+});
+
 // src/vendors/catalog/index.ts
 var ALL_MODELS = [
   ...MODELS,
@@ -8518,7 +8598,8 @@ var ALL_MODELS = [
   ...MODELS33,
   ...MODELS34,
   ...MODELS35,
-  ...MODELS36
+  ...MODELS36,
+  ...MODELS37
 ];
 var getModelsByMode = (mode, includeDisabled = false) => ALL_MODELS.filter((m) => m.mode === mode && (includeDisabled || isVisibleForReleases(m)));
 
@@ -10610,6 +10691,7 @@ var AsyncFlashV1 = "async-flash-v1";
 var BytedanceOmnihumanV15 = "bytedance-omnihuman-v1.5";
 var BytedanceVideoEnhance = "bytedance-video-enhance";
 var BytedanceVideoUpscaler = "bytedance-video-upscaler";
+var CaptionsaiVideoCaptions = "captionsai-video-captions";
 var ClaudeHaiku45 = "claude-haiku-4-5";
 var ClaudeOpus48 = "claude-opus-4-8";
 var ClaudeSonnet46 = "claude-sonnet-4-6";
@@ -10821,6 +10903,7 @@ var Models = {
   BytedanceOmnihumanV15,
   BytedanceVideoEnhance,
   BytedanceVideoUpscaler,
+  CaptionsaiVideoCaptions,
   ClaudeHaiku45,
   ClaudeOpus48,
   ClaudeSonnet46,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@picsart/ai-sdk",
-  "version": "5.22.1",
+  "version": "5.24.0",
   "type": "module",
   "description": "Type-safe SDK for 100+ AI models — image, video, audio, and text generation with Picsart",
   "license": "MIT",