@picsart/ai-sdk 5.19.1 → 5.21.0

package/index.d.ts

@@ -227,6 +227,13 @@ type ModelInputById = {
         count?: 1 | 2 | 4 | 6 | 8 | 10;
         imageUrls?: string[];
     };
+    "flux-video-upscale": {
+        videoUrl: string;
+        upscaleFactor?: number;
+        creativity?: 0 | 1;
+        prompt?: string;
+        safetyTolerance?: number;
+    };
     "gemini-2.5-flash-image": {
         prompt: string;
         aspectRatio?: "1:1" | "16:9" | "9:16" | "3:4" | "4:3" | "2:3" | "21:9";
@@ -2180,9 +2187,10 @@ interface ModelDefinition {
     editEstimatedTime?: number | Record<string, number>;
     testTimeout?: number;
     /**
-     * Per-model polling overrides for async jobs. Widens the global
-     * 2s × 300-attempt (~10 min) default for models whose generations can
-     * outlast it. Explicit per-call poll options still win.
+     * Per-model polling overrides for async jobs. Narrows or widens the mode
+     * default (video 2s × 1800 ≈ 1 h; image/audio/text 1s × 1200 ≈ 20 min) for
+     * models whose generations don't fit it. Explicit per-call poll options
+     * still win.
      */
     pollOptions?: {
         intervalMs?: number;
@@ -2389,6 +2397,17 @@ interface GenerateTextResult {
 /** Options for individual generate() / submit() calls. */
 interface GenerateOptions {
     signal?: AbortSignal;
+    /**
+     * Poll interval for the async status loop, in ms. Overrides the model's
+     * `pollOptions` and the mode default (video 2s; image/audio/text 1s).
+     */
+    intervalMs?: number;
+    /**
+     * Max poll attempts before the call throws a timeout. Overrides the model's
+     * `pollOptions` and the mode default (video 1800 ≈ 1 hour;
+     * image/audio/text 1200 ≈ 20 min).
+     */
+    maxAttempts?: number;
     /** Save to a specific subfolder instead of the root (legacy — used by SDK DriveConfig). */
     folder?: DriveFolder;
     /** Save result to Picsart Drive via backend. Injected into the workflow payload. */
@@ -2493,6 +2512,7 @@ declare const Models: {
     readonly Flux3Video: "flux-3-video";
     readonly FluxKontextMax: "flux-kontext-max";
     readonly FluxKontextPro: "flux-kontext-pro";
+    readonly FluxVideoUpscale: "flux-video-upscale";
     readonly Gemini25FlashImage: "gemini-2.5-flash-image";
     readonly Gemini25FlashTts: "gemini-2.5-flash-tts";
     readonly Gemini25ProTts: "gemini-2.5-pro-tts";

package/index.js

@@ -3129,9 +3129,6 @@ var { MODELS: MODELS10 } = defineModels("wan", [
     description: "Wan 3.0 all-in-one \u2014 text, image/video/audio references, and start/end frames with adaptive ratio, intelligent duration, and audio.",
     features: [...wanV3Features],
     constraints: wanV3Constraints,
-    // Generations at 1080P / long durations can outlast the global 10-min
-    // polling default — widen to 5s × 360 attempts (30 min).
-    pollOptions: { intervalMs: 5e3, maxAttempts: 360 },
     paramConfig: { ...wanV3ParamConfig }
   },
   {
@@ -3146,7 +3143,6 @@ var { MODELS: MODELS10 } = defineModels("wan", [
     description: "Wan 3.0 Prime \u2014 the same all-in-one model as Wan 3.0, up to 7x faster.",
     features: [...wanV3Features],
     constraints: wanV3Constraints,
-    pollOptions: { intervalMs: 5e3, maxAttempts: 360 },
     paramConfig: { ...wanV3ParamConfig }
   }
 ]);
@@ -3786,12 +3782,6 @@ var { MODELS: MODELS12 } = defineModels("seedance", [
     badge: ["new", "premium", "hot"],
     description: "Latest cinematic video with audio, multi-reference input, and mp4/mov output. Up to 30s.",
     features: [feat("Reference Image", "frame"), feat("Start/End Frame", "frame"), feat("Audio", "audio"), feat("1080p", "resolution"), feat("4-30 sec", "duration")],
-    // 1080p / 30s runs, and the whole extend path, outlast the global 10-min
-    // polling default: measured p90 for seedance-2.5-video-extend sits above
-    // the 600s ceiling. Widened to 5s × 360 attempts (30 min), same request
-    // count, 3x the wall clock. Deliberately NOT derived from estimatedTime,
-    // which understates seedance latency by ~10x.
-    pollOptions: { intervalMs: 5e3, maxAttempts: 360 },
     paramConfig: {
       ...params.prompt(),
       ...params.aspectRatio(SEEDANCE_AR),
@@ -3821,7 +3811,6 @@ var { MODELS: MODELS12 } = defineModels("seedance", [
     badge: ["new", "premium", "hot"],
     description: "Edit video \u2014 replace subjects, add or remove objects, restyle scenes with reference images.",
     features: [feat("Video Input", "input"), feat("Multi-Image Input", "input"), feat("Audio", "audio"), feat("1080p", "resolution"), feat("Source length", "duration")],
-    pollOptions: { intervalMs: 5e3, maxAttempts: 360 },
     paramConfig: {
       ...params.prompt(),
       // Editing mode: aspect ratio is fixed to 'adaptive' and duration is
@@ -3848,7 +3837,6 @@ var { MODELS: MODELS12 } = defineModels("seedance", [
     badge: ["new", "premium", "hot"],
     description: "Stitch up to 10 clips into one continuous, extended video.",
     features: [feat("Multi-Video Input", "input"), feat("Audio", "audio"), feat("1080p", "resolution"), feat("4-30 sec", "duration")],
-    pollOptions: { intervalMs: 5e3, maxAttempts: 360 },
     paramConfig: {
       ...params.prompt(),
       // Extension mode: aspect ratio is locked to 'adaptive' (vendor rule);
@@ -5506,6 +5494,36 @@ var { MODELS: MODELS21 } = defineModels("flux", [
       // draft: fast low-step preview.
       ...p.boolean("draft", false, "Draft")
     }
+  },
+  {
+    // Pure pass-through: the worker Command takes the SDK's own field names
+    // (videoUrl, upscaleFactor, creativity, prompt, safetyTolerance) — no
+    // payload builder needed. webhookUrl is polling plumbing, not surfaced
+    // (same as flux-3-video).
+    id: "flux-video-upscale",
+    name: "Flux Video Upscale",
+    workflow: "flux/v1/video-upscale",
+    mode: "video",
+    inputType: "v2v",
+    addedAt: "2026-08-28",
+    estimatedTime: 180,
+    description: "Upscale videos toward 4K (1.5x\u20133x) in precise (source-faithful) or creative (detail-enhancing) mode. Source clips up to 20 seconds and 2K.",
+    features: [
+      feat("Upscale", "quality"),
+      feat("Video Required", "input"),
+      feat("Up to 4K", "resolution")
+    ],
+    paramConfig: {
+      // Vendor source caps: 20s, 50 MB, 2560x1440 (2K).
+      ...params.videoInput("Source Video", "asset", true, 20, 1440, 50 * 1024 * 1024),
+      ...p.range("upscaleFactor", 1.5, 3, 2, { step: 0.5, label: "Upscale Factor" }),
+      // Vendor switch: 0 preserves the source precisely; 1 (default) allows
+      // creative detail enhancement and is the more expensive pricing tier.
+      ...p.enum("creativity", [{ id: 0, label: "Precise" }, { id: 1, label: "Creative" }], 1, { label: "Creativity" }),
+      ...params.prompt({ required: false }),
+      // Moderation level: 0 (strict) … 4 (permissive).
+      ...p.range("safetyTolerance", 0, 4, 2, { label: "Safety Tolerance" })
+    }
   }
 ]);
 
@@ -9874,6 +9892,23 @@ function createCatalogs(transport, options) {
 }
 
 // src/client/index.ts
+var MODE_POLL_DEFAULTS = {
+  video: { intervalMs: 2e3, maxAttempts: 1800 },
+  // 2s × 1800 = 1 hour
+  image: { intervalMs: 1e3, maxAttempts: 1200 },
+  // 1s × 1200 = 20 min
+  audio: { intervalMs: 1e3, maxAttempts: 1200 },
+  // 1s × 1200 = 20 min
+  text: { intervalMs: 1e3, maxAttempts: 1200 }
+  // 1s × 1200 = 20 min
+};
+function resolvePollOptions(model, overrides) {
+  const resolved = { ...MODE_POLL_DEFAULTS[model.mode], ...model.pollOptions };
+  if (overrides?.intervalMs !== void 0) resolved.intervalMs = overrides.intervalMs;
+  if (overrides?.maxAttempts !== void 0) resolved.maxAttempts = overrides.maxAttempts;
+  if (overrides?.signal !== void 0) resolved.signal = overrides.signal;
+  return resolved;
+}
 function createClient(config) {
   const isConfig = isClientConfig(config);
   const transport = isConfig ? buildTransport(config) : config;
@@ -9883,7 +9918,8 @@ function createClient(config) {
   const catalogs = createCatalogs(transport, isConfig ? config.catalogs : void 0);
   const driveConfig = isConfig ? config.drive : void 0;
   const driveClient = isConfig && driveConfig ? createDriveClient(resolveFetch(config), config.apiUrl, driveConfig.folder) : null;
-  async function executeModel(model, workflow, payload, signal) {
+  async function executeModel(model, workflow, payload, options) {
+    const signal = options?.signal;
     if (model.syncExecute || !supportsSubmit) {
       const syncResponse = await client.run(
         { workflow, payload, signal },
@@ -9896,7 +9932,7 @@ function createClient(config) {
         syncResponse.usage
       );
     }
-    return client.run({ workflow, payload, signal }, model.pollOptions);
+    return client.run({ workflow, payload, signal }, resolvePollOptions(model, options));
   }
   function buildDrivePayloadOptions(model, params2, options) {
     const explicit = options?.drive;
@@ -9936,7 +9972,7 @@ function createClient(config) {
       const { workflow, payload, contract } = prepareRequest(resolved, params2);
       const drive = buildDrivePayloadOptions(resolved, params2, options);
       const finalPayload = injectDriveOptions(payload, drive);
-      const completed = await executeModel(resolved, workflow, finalPayload, options?.signal);
+      const completed = await executeModel(resolved, workflow, finalPayload, options);
       return parseResult(completed, resolved, contract);
     },
     /**
@@ -9953,7 +9989,7 @@ function createClient(config) {
         throw new Error(`${resolved.name} is not a text model \u2014 use generate() instead.`);
       }
       const { workflow, payload } = prepareRequest(resolved, params2);
-      const completed = await executeModel(resolved, workflow, payload, options?.signal);
+      const completed = await executeModel(resolved, workflow, payload, options);
       return parseTextResult(completed, resolved);
     },
     /** @deprecated Use `getCredits()` instead. */
@@ -9997,7 +10033,7 @@ function createClient(config) {
     async result(handle, model, options) {
       const resolved = resolveModel(model);
       const contract = getModelContract(resolved.id);
-      const completed = await client.result(handle, { ...resolved.pollOptions, ...options });
+      const completed = await client.result(handle, resolvePollOptions(resolved, options));
       return parseResult(completed, resolved, contract);
     },
     /**
@@ -10510,6 +10546,7 @@ var Flux2Pro = "flux-2-pro";
 var Flux3Video = "flux-3-video";
 var FluxKontextMax = "flux-kontext-max";
 var FluxKontextPro = "flux-kontext-pro";
+var FluxVideoUpscale = "flux-video-upscale";
 var Gemini25FlashImage = "gemini-2.5-flash-image";
 var Gemini25FlashTts = "gemini-2.5-flash-tts";
 var Gemini25ProTts = "gemini-2.5-pro-tts";
@@ -10718,6 +10755,7 @@ var Models = {
   Flux3Video,
   FluxKontextMax,
   FluxKontextPro,
+  FluxVideoUpscale,
   Gemini25FlashImage,
   Gemini25FlashTts,
   Gemini25ProTts,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@picsart/ai-sdk",
-  "version": "5.19.1",
+  "version": "5.21.0",
   "type": "module",
   "description": "Type-safe SDK for 100+ AI models — image, video, audio, and text generation with Picsart",
   "license": "MIT",