@picsart/ai-sdk 5.17.0 → 5.19.0

package/index.d.ts

@@ -288,6 +288,17 @@ type ModelInputById = {
         imageUrls?: string[];
         thinking?: "off" | "low" | "medium" | "high";
     };
+    "gemini-omni-1.1-flash-preview": {
+        prompt: string;
+        aspectRatio?: "16:9" | "9:16";
+        resolution?: "360p" | "720p" | "1080p" | "4k";
+        duration?: 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10;
+        startFrame?: string;
+        endFrame?: string;
+        imageUrls?: string[];
+        videoUrl?: string;
+        videoUrls?: string[];
+    };
     "gemini-omni-flash-preview": {
         prompt: string;
         aspectRatio?: "16:9" | "9:16";
@@ -1516,6 +1527,21 @@ type ModelInputById = {
         watermark?: boolean;
         seed?: number;
     };
+    "wan-3.0-video-prime": {
+        prompt: string;
+        duration?: 5 | 10 | 15 | 30;
+        resolution?: "480P" | "720P" | "1080P";
+        aspectRatio?: "16:9" | "9:16" | "1:1" | "4:3" | "3:4" | "adaptive";
+        generateAudio?: boolean;
+        startFrame?: string;
+        endFrame?: string;
+        imageUrls?: string[];
+        videoUrls?: string[];
+        audioUrls?: string[];
+        enableThinking?: boolean;
+        watermark?: boolean;
+        seed?: number;
+    };
 };
 type TypedModelId = keyof ModelInputById;
 type ModelInput<M extends TypedModelId> = ModelInputById[M];
@@ -2153,6 +2179,15 @@ interface ModelDefinition {
     estimatedTime?: number | Record<string, number>;
     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.
+     */
+    pollOptions?: {
+        intervalMs?: number;
+        maxAttempts?: number;
+    };
 }
 
 /**
@@ -2468,6 +2503,7 @@ declare const Models: {
     readonly Gemini35FlashLite: "gemini-3.5-flash-lite";
     readonly Gemini36Flash: "gemini-3.6-flash";
     readonly Gemini37Flash: "gemini-3.7-flash";
+    readonly GeminiOmni11FlashPreview: "gemini-omni-1.1-flash-preview";
     readonly GeminiOmniFlashPreview: "gemini-omni-flash-preview";
     readonly Gpt55: "gpt-5.5";
     readonly GptImage1: "gpt-image-1";
@@ -2637,6 +2673,7 @@ declare const Models: {
     readonly Wan27T2v: "wan-2.7-t2v";
     readonly Wan27VideoEdit: "wan-2.7-video-edit";
     readonly Wan30Video: "wan-3.0-video";
+    readonly Wan30VideoPrime: "wan-3.0-video-prime";
     /** @deprecated Use the `catalog` accessor (`catalog.all()` / `catalog.find({ output, provider })`) instead. */
     readonly list: (filter?: ModelFilter) => ModelDefinition[];
     /** @deprecated Use `Model(id).validate(input)` instead. */

package/index.js

@@ -1201,6 +1201,7 @@ function defineModels(provider, configs) {
     if (c.estimatedTime !== void 0) model.estimatedTime = c.estimatedTime;
     if (c.editEstimatedTime !== void 0) model.editEstimatedTime = c.editEstimatedTime;
     if (c.testTimeout !== void 0) model.testTimeout = c.testTimeout;
+    if (c.pollOptions !== void 0) model.pollOptions = c.pollOptions;
     if (c.badge !== void 0) model.badge = c.badge;
     if (c.addedAt !== void 0) model.addedAt = c.addedAt;
     if (c.disabled !== void 0) model.disabled = c.disabled;
@@ -2927,6 +2928,22 @@ var wanV3Constraints = [
     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 = {
+  ...params.prompt(),
+  ...params.duration([5, 10, 15, 30], 5),
+  ...params.resolution(["480P", "720P", "1080P"], "1080P"),
+  ...params.aspectRatio(["16:9", "9:16", "1:1", "4:3", "3:4", "adaptive"]),
+  ...params.generateAudio(true),
+  ...params.startFrame(),
+  ...params.endFrame(),
+  ...params.imageInput(10, "Reference Images"),
+  ...params.videoInputs(5, "Reference Videos", false),
+  ...params.audioInputs(5, "Reference Audios"),
+  ...p.boolean("enableThinking", false, "Deep Thinking"),
+  ...p.boolean("watermark", false, "Watermark"),
+  ...p.range("seed", 0, 2147483647, 0, { label: "Seed" })
+};
 var { MODELS: MODELS10 } = defineModels("wan", [
   // ── Video ─────────────────────────────────────────
   {
@@ -3094,6 +3111,10 @@ var { MODELS: MODELS10 } = defineModels("wan", [
     }
   },
   // ── Wan 3.0 all-in-one Video ─────────────────────────
+  // wan-3.0-video and wan-3.0-video-prime share the wan/v3/video workflow and
+  // the full param surface — the backend `model` enum value (hardcoded per
+  // entry in wan.payloads.ts) is the only wire difference. Prime is the same
+  // model, up to 7x faster.
   {
     id: "wan-3.0-video",
     name: "Wan 3.0",
@@ -3106,28 +3127,32 @@ var { MODELS: MODELS10 } = defineModels("wan", [
     mode: "video",
     inputType: "t2v",
     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: [feat("Image Input", "input"), feat("Video Input", "input"), feat("Audio", "audio"), feat("Start/End Frame", "frame"), feat("1080P", "resolution"), feat("Adaptive Ratio", "resolution")],
+    features: [...wanV3Features],
     constraints: wanV3Constraints,
-    paramConfig: {
-      ...params.prompt(),
-      ...params.duration([5, 10, 15, 30], 5),
-      ...params.resolution(["480P", "720P", "1080P"], "1080P"),
-      ...params.aspectRatio(["16:9", "9:16", "1:1", "4:3", "3:4", "adaptive"]),
-      ...params.generateAudio(true),
-      ...params.startFrame(),
-      ...params.endFrame(),
-      ...params.imageInput(10, "Reference Images"),
-      ...params.videoInputs(5, "Reference Videos", false),
-      ...params.audioInputs(5, "Reference Audios"),
-      ...p.boolean("enableThinking", false, "Deep Thinking"),
-      ...p.boolean("watermark", false, "Watermark"),
-      ...p.range("seed", 0, 2147483647, 0, { label: "Seed" })
-    }
+    // 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 }
+  },
+  {
+    id: "wan-3.0-video-prime",
+    name: "Wan 3.0 Prime",
+    modelId: "wan3.0-video-prime",
+    addedAt: "2026-08-26",
+    workflow: "wan/v3/video",
+    estimatedTime: 30,
+    mode: "video",
+    inputType: "t2v",
+    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 }
   }
 ]);
 
 // src/vendors/catalog/wan.payloads.ts
-var buildWanV3VideoPayload = (input) => {
+var makeWanV3VideoPayload = (model) => (input) => {
   const media = [];
   if (input.startFrame) media.push({ type: "first_frame", url: input.startFrame });
   if (input.endFrame) media.push({ type: "last_frame", url: input.endFrame });
@@ -3141,7 +3166,7 @@ var buildWanV3VideoPayload = (input) => {
     for (const url of input.audioUrls) media.push({ type: "reference_audio", url });
   }
   return {
-    model: "wan3.0-video",
+    model,
     resolution: input.resolution ?? "1080P",
     ratio: input.aspectRatio ?? "16:9",
     duration: input.duration ?? 5,
@@ -3154,7 +3179,8 @@ var buildWanV3VideoPayload = (input) => {
   };
 };
 registerPayloads(MODELS10, {
-  "wan-3.0-video": buildWanV3VideoPayload
+  "wan-3.0-video": makeWanV3VideoPayload("wan3.0-video"),
+  "wan-3.0-video-prime": makeWanV3VideoPayload("wan3.0-video-prime")
 });
 
 // src/vendors/catalog/luma.ts
@@ -5727,9 +5753,56 @@ var { MODELS: MODELS22 } = defineModels("google", [
       ...params.imageInput(1, "Source Image", false, "asset"),
       ...params.videoInput("Source Video", "asset", false)
     }
+  },
+  {
+    id: "gemini-omni-1.1-flash-preview",
+    name: "Gemini Omni 1.2 Flash",
+    specName: "Gemini Omni 1.1 Flash Preview",
+    addedAt: "2026-08-27",
+    workflow: "gemini-omni/video",
+    estimatedTime: { "360p": 30, "720p": 40, "1080p": 60, "4k": 90 },
+    mode: "video",
+    inputType: "t2v",
+    description: "Gemini Omni with frame interpolation, video extension, reference-guided generation, and up to 4K output.",
+    features: [feat("Start/End Frame", "input"), feat("Reference Images & Videos", "input"), feat("Video Extension", "input"), feat("4K", "resolution"), feat("3\u201310 sec", "duration")],
+    paramConfig: {
+      ...params.prompt(),
+      ...params.aspectRatio(["16:9", "9:16"], "16:9"),
+      ...params.resolution(["360p", "720p", "1080p", "4k"], "720p"),
+      ...params.duration([3, 4, 5, 6, 7, 8, 9, 10], 8),
+      ...params.startFrame("Start Frame"),
+      ...params.endFrame("End Frame"),
+      ...params.imageInput(5, "Reference Images", false, "reference"),
+      // Extension source: the worker extends the clip by up to 10s; input must be under 30s.
+      ...params.videoInput("Source Video", "asset", false, 30),
+      ...params.videoInputs(3, "Reference Videos")
+    }
   }
 ]);
 
+// src/vendors/catalog/gemini.payloads.ts
+function inferMimeType3(url) {
+  return url.match(/\.png(\?|$)/i) ? "image/png" : "image/jpeg";
+}
+var toImage = (url) => ({ url, mimeType: inferMimeType3(url) });
+var buildOmniFlash11Payload = (input) => ({
+  prompt: input.prompt,
+  model: "gemini-omni-1.1-flash-preview",
+  // Materialize the catalog defaults so direct SDK calls send the advertised
+  // values rather than relying on the worker/vendor defaults.
+  aspectRatio: input.aspectRatio ?? "16:9",
+  durationSeconds: input.duration ?? 8,
+  resolution: input.resolution ?? "720p",
+  ...input.startFrame ? { image: toImage(input.startFrame) } : {},
+  ...input.endFrame ? { lastFrame: toImage(input.endFrame) } : {},
+  ...input.imageUrls?.length ? { referenceImages: input.imageUrls.map(toImage) } : {},
+  ...input.videoUrl ? { video: { url: input.videoUrl } } : {},
+  ...input.videoUrls?.length ? { referenceVideos: input.videoUrls.map((url) => ({ url })) } : {}
+});
+registerPayloads(MODELS22, {
+  "gemini-omni-1.1-flash-preview": buildOmniFlash11Payload
+});
+
 // src/vendors/catalog/openai.ts
 var GPT_IMAGE_AR_TO_SIZE = {
   "1:1": "1024x1024",
@@ -9815,7 +9888,7 @@ function createClient(config) {
         syncResponse.usage
       );
     }
-    return client.run({ workflow, payload, signal });
+    return client.run({ workflow, payload, signal }, model.pollOptions);
   }
   function buildDrivePayloadOptions(model, params2, options) {
     const explicit = options?.drive;
@@ -9916,7 +9989,7 @@ function createClient(config) {
     async result(handle, model, options) {
       const resolved = resolveModel(model);
       const contract = getModelContract(resolved.id);
-      const completed = await client.result(handle, options);
+      const completed = await client.result(handle, { ...resolved.pollOptions, ...options });
       return parseResult(completed, resolved, contract);
     },
     /**
@@ -10439,6 +10512,7 @@ var Gemini31FlashLiteImage = "gemini-3.1-flash-lite-image";
 var Gemini35FlashLite = "gemini-3.5-flash-lite";
 var Gemini36Flash = "gemini-3.6-flash";
 var Gemini37Flash = "gemini-3.7-flash";
+var GeminiOmni11FlashPreview = "gemini-omni-1.1-flash-preview";
 var GeminiOmniFlashPreview = "gemini-omni-flash-preview";
 var Gpt55 = "gpt-5.5";
 var GptImage1 = "gpt-image-1";
@@ -10608,6 +10682,7 @@ var Wan27R2v = "wan-2.7-r2v";
 var Wan27T2v = "wan-2.7-t2v";
 var Wan27VideoEdit = "wan-2.7-video-edit";
 var Wan30Video = "wan-3.0-video";
+var Wan30VideoPrime = "wan-3.0-video-prime";
 var Models = {
   AsyncFlashV1,
   BytedanceOmnihumanV15,
@@ -10645,6 +10720,7 @@ var Models = {
   Gemini35FlashLite,
   Gemini36Flash,
   Gemini37Flash,
+  GeminiOmni11FlashPreview,
   GeminiOmniFlashPreview,
   Gpt55,
   GptImage1,
@@ -10814,6 +10890,7 @@ var Models = {
   Wan27T2v,
   Wan27VideoEdit,
   Wan30Video,
+  Wan30VideoPrime,
   /** @deprecated Use the `catalog` accessor (`catalog.all()` / `catalog.find({ output, provider })`) instead. */
   list(filter) {
     if (!filter) return [...ALL_MODELS];

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@picsart/ai-sdk",
-  "version": "5.17.0",
+  "version": "5.19.0",
   "type": "module",
   "description": "Type-safe SDK for 100+ AI models — image, video, audio, and text generation with Picsart",
   "license": "MIT",