@picsart/ai-sdk 5.22.0 → 5.23.0

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[];
@@ -968,6 +972,17 @@ type ModelInputById = {
         enhancePrompt?: boolean;
         imageUrls?: string[];
         promptExtendMode?: "direct" | "agent";
+        enableThinking?: boolean;
+    };
+    "qwen-image-3.0-pro": {
+        prompt: string;
+        negativePrompt?: string;
+        resolution?: "2048x2048" | "2688x1536" | "1536x2688" | "2368x1728" | "1728x2368";
+        count?: 1 | 2 | 4 | 6;
+        enhancePrompt?: boolean;
+        imageUrls?: string[];
+        promptExtendMode?: "direct" | "agent";
+        enableThinking?: boolean;
     };
     "recraft-creative-upscale": {
         imageUrls: [string, ...string[]];
@@ -2002,7 +2017,7 @@ 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. */
@@ -2100,8 +2115,10 @@ interface GenerationContext {
     substyle?: string;
     thinkingLevel?: 'minimal' | 'high';
     thinkingBudget?: number;
-    /** Qwen 3.0 — prompt-rewrite strategy (`direct`/`agent`), sent as `prompt_extend_mode`. */
+    /** Qwen 3.0 family — prompt-rewrite strategy (`direct`/`agent`), sent as `prompt_extend_mode`. */
     promptExtendMode?: 'direct' | 'agent';
+    /** Qwen 3.0 family — thinking mode (requires prompt_extend), sent as `enable_thinking`. */
+    enableThinking?: boolean;
 }
 type PayloadBuilder<TContext extends GenerationContext = GenerationContext> = (ctx: TContext) => object;
 /** Lightweight runtime schema contract used by SDK integrations. */
@@ -2501,6 +2518,7 @@ 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";
@@ -2630,6 +2648,7 @@ declare const Models: {
     readonly QwenImage2: "qwen-image-2";
     readonly QwenImage2Pro: "qwen-image-2-pro";
     readonly QwenImage30: "qwen-image-3.0";
+    readonly QwenImage30Pro: "qwen-image-3.0-pro";
     readonly RecraftCreativeUpscale: "recraft-creative-upscale";
     readonly RecraftCrispUpscale: "recraft-crisp-upscale";
     readonly RecraftExplore: "recraft-explore";

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
@@ -1175,7 +1176,7 @@ 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 +1212,19 @@ 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;
   }
 }
@@ -6735,6 +6736,7 @@ var buildQwen2Payload = (ctx) => {
 };
 var buildQwenV1 = (model) => (ctx) => {
   const hasImages = Array.isArray(ctx.imageUrls) && ctx.imageUrls.length > 0;
+  const promptExtendMode = ctx.promptExtendMode === "agent" && hasImages && model === "qwen-image-3.0-pro" ? void 0 : ctx.promptExtendMode;
   return {
     prompt: ctx.prompt,
     model,
@@ -6743,8 +6745,10 @@ var buildQwenV1 = (model) => (ctx) => {
     size: (ctx.resolution ?? "2048x2048").replace("x", "*"),
     n: ctx.count ?? 1,
     prompt_extend: ctx.enhancePrompt ?? true,
-    // Qwen 3.0 only — prompt-rewrite strategy (direct/agent); 2.x ignores it.
-    ...ctx.promptExtendMode ? { prompt_extend_mode: ctx.promptExtendMode } : {},
+    // Qwen 3.0 family only — prompt-rewrite strategy (direct/agent); 2.x ignores it.
+    ...promptExtendMode ? { prompt_extend_mode: promptExtendMode } : {},
+    // Qwen 3.0 family only — thinking mode (requires prompt_extend).
+    ...ctx.enableThinking != null ? { enable_thinking: ctx.enableThinking } : {},
     watermark: false,
     ...ctx.seed != null ? { seed: ctx.seed } : {}
   };
@@ -6759,7 +6763,8 @@ var qwenV1Params = {
 };
 var qwenV1Params3 = {
   ...qwenV1Params,
-  ...p.enum("promptExtendMode", ["direct", "agent"], "direct")
+  ...p.enum("promptExtendMode", ["direct", "agent"], "direct"),
+  ...p.boolean("enableThinking", true, "Deep Thinking")
 };
 var { MODELS: MODELS28 } = defineModels("qwen", [
   {
@@ -6837,6 +6842,25 @@ var { MODELS: MODELS28 } = defineModels("qwen", [
       feat("2K", "resolution")
     ],
     paramConfig: qwenV1Params3
+  },
+  {
+    id: "qwen-image-3.0-pro",
+    name: "Qwen 3.0 Pro",
+    addedAt: "2026-08-31",
+    workflow: "qwen/v1/text-to-image",
+    editWorkflow: "qwen/v1/image-to-image",
+    buildPayload: buildQwenV1("qwen-image-3.0-pro"),
+    estimatedTime: 90,
+    mode: "image",
+    inputType: "t2i",
+    badge: ["premium"],
+    description: "Qwen-Image 3.0 Pro (GA) \u2014 flagship text-to-image and image editing with prompt-rewrite modes and thinking mode.",
+    features: [
+      feat("Image Input", "input"),
+      feat("Negative Prompt", "characteristic"),
+      feat("2K", "resolution")
+    ],
+    paramConfig: qwenV1Params3
   }
 ]);
 
@@ -8458,6 +8482,54 @@ 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,
@@ -8495,7 +8567,8 @@ var ALL_MODELS = [
   ...MODELS33,
   ...MODELS34,
   ...MODELS35,
-  ...MODELS36
+  ...MODELS36,
+  ...MODELS37
 ];
 var getModelsByMode = (mode, includeDisabled = false) => ALL_MODELS.filter((m) => m.mode === mode && (includeDisabled || isVisibleForReleases(m)));
 
@@ -10587,6 +10660,7 @@ 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";
@@ -10716,6 +10790,7 @@ var Qwen = "qwen";
 var QwenImage2 = "qwen-image-2";
 var QwenImage2Pro = "qwen-image-2-pro";
 var QwenImage30 = "qwen-image-3.0";
+var QwenImage30Pro = "qwen-image-3.0-pro";
 var RecraftCreativeUpscale = "recraft-creative-upscale";
 var RecraftCrispUpscale = "recraft-crisp-upscale";
 var RecraftExplore = "recraft-explore";
@@ -10797,6 +10872,7 @@ var Models = {
   BytedanceOmnihumanV15,
   BytedanceVideoEnhance,
   BytedanceVideoUpscaler,
+  CaptionsaiVideoCaptions,
   ClaudeHaiku45,
   ClaudeOpus48,
   ClaudeSonnet46,
@@ -10926,6 +11002,7 @@ var Models = {
   QwenImage2,
   QwenImage2Pro,
   QwenImage30,
+  QwenImage30Pro,
   RecraftCreativeUpscale,
   RecraftCrispUpscale,
   RecraftExplore,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@picsart/ai-sdk",
-  "version": "5.22.0",
+  "version": "5.23.0",
   "type": "module",
   "description": "Type-safe SDK for 100+ AI models — image, video, audio, and text generation with Picsart",
   "license": "MIT",