@picsart/ai-sdk 6.48.1 → 6.49.0

package/index.d.ts

@@ -1234,6 +1234,18 @@ type ModelInputById = {
         prompt: string;
         imageUrls?: string[];
     };
+    "macaron-cookie": {
+        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[];
+        task?: "edit" | "extend";
+    };
     "minimax-h3": {
         prompt: string;
         startFrame?: string;
@@ -3720,6 +3732,7 @@ declare const Models: {
     readonly Lyria3Clip: "lyria-3-clip";
     readonly Lyria3Pro: "lyria-3-pro";
     readonly Lyria35: "lyria-3.5";
+    readonly MacaronCookie: "macaron-cookie";
     readonly MinimaxH3: "minimax-h3";
     readonly MinimaxH3ContextIr: "minimax-h3-context-ir";
     readonly MinimaxH3Max: "minimax-h3-max";

package/index.js

@@ -7252,6 +7252,41 @@ var { MODELS: MODELS22 } = defineModels("google", [
       ...params.videoInput("Source Video", "asset", false, 30),
       ...params.videoInputs(3, "Reference Videos")
     }
+  },
+  {
+    id: "macaron-cookie",
+    name: "Gemini Omni (Macaron)",
+    addedAt: "2026-10-19",
+    release: "preview",
+    workflow: "gemini-omni/video",
+    estimatedTime: { "360p": 30, "720p": 40, "1080p": 60, "4k": 90 },
+    mode: "video",
+    inputType: "t2v",
+    description: "Gemini Omni (Macaron) \u2014 multimodal video with frame interpolation, 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"),
+      // Source video drives the edit/extend tasks; input must be under 30s.
+      ...params.videoInput("Source Video", "asset", false, 30),
+      ...params.videoInputs(3, "Reference Videos"),
+      // With a source video the worker auto-derives `extend`; this toggle lets
+      // the caller choose video editing instead. Only sent when a video is present.
+      ...p.enum("task", ["edit", "extend"], "edit", { label: "Mode" })
+    },
+    // edit/extend only apply to a source video — disable the toggle until one
+    // is provided (matches the payload builder, which omits `task` without it).
+    constraints: [
+      {
+        when: { videoUrl: { exists: false } },
+        then: { task: { disabled: true, reason: "Add a source video to choose edit or extend." } }
+      }
+    ]
   }
 ]);
 
@@ -7274,6 +7309,21 @@ var buildOmniFlash11Payload = (input) => ({
   ...input.videoUrl ? { video: { url: input.videoUrl } } : {},
   ...input.videoUrls?.length ? { referenceVideos: input.videoUrls.map((url) => ({ url })) } : {}
 });
+var buildMacaronCookiePayload = (input) => ({
+  prompt: input.prompt,
+  model: "macaron-cookie",
+  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) } : {},
+  // A source video runs edit or extend; the worker auto-derives `extend`, so
+  // forward the explicit task (default 'edit') to pick editing. Omitted without
+  // a video so text/image/reference inputs keep their own derived tasks.
+  ...input.videoUrl ? { video: { url: input.videoUrl }, task: input.task ?? "edit" } : {},
+  ...input.videoUrls?.length ? { referenceVideos: input.videoUrls.map((url) => ({ url })) } : {}
+});
 var STYLE_MAX = 500;
 var TEXT_MAX = 6e3;
 var invalid = (message) => new ApiError(message, { status: 400, code: "validation_error" });
@@ -7359,6 +7409,7 @@ var buildTts38Payload = (model) => (input) => {
 };
 registerPayloads(MODELS22, {
   "gemini-omni-1.1-flash-preview": buildOmniFlash11Payload,
+  "macaron-cookie": buildMacaronCookiePayload,
   "gemini-2.5-flash-tts": buildTts25Payload("gemini-2.5-flash-tts"),
   "gemini-2.5-pro-tts": buildTts25Payload("gemini-2.5-pro-tts"),
   "gemini-3.8-flash-tts": buildTts38Payload("gemini-3.8-flash-tts"),
@@ -11878,13 +11929,12 @@ var extractUrl = (result) => {
     if (typeof obj.url === "string") return obj.url;
     if (typeof obj.videoUrl === "string") return obj.videoUrl;
     if (typeof obj.video_url === "string") return obj.video_url;
-    if (obj.video && typeof obj.video === "object") {
-      const v = obj.video;
-      if (typeof v.url === "string") return v.url;
-    }
-    if (obj.audio && typeof obj.audio === "object") {
-      const a = obj.audio;
-      if (typeof a.url === "string") return a.url;
+    for (const key of ["video", "audio"]) {
+      const media = Array.isArray(obj[key]) ? obj[key][0] : obj[key];
+      if (media && typeof media === "object") {
+        const m = media;
+        if (typeof m.url === "string") return m.url;
+      }
     }
     if (obj.assets && typeof obj.assets === "object") {
       const a = obj.assets;
@@ -11995,7 +12045,7 @@ var extractText = (result) => {
   }
   return void 0;
 };
-var RESULT_ARRAY_KEYS = ["items", "images", "imageUrls", "urls", "data", "previews"];
+var RESULT_ARRAY_KEYS = ["items", "images", "imageUrls", "urls", "data", "previews", "audio", "video"];
 var extractAllResults = (result) => {
   if (!result || typeof result !== "object") return void 0;
   const obj = result;
@@ -13836,7 +13886,7 @@ function parseResult(completed, model, contract) {
   }
   const obj = parsed && typeof parsed === "object" ? parsed : void 0;
   let source = parsed;
-  for (const key of ["images", "items", "imageUrls", "data", "previews"]) {
+  for (const key of ["images", "items", "imageUrls", "data", "previews", "audio", "video"]) {
     const arr = obj?.[key];
     if (Array.isArray(arr) && arr.length > 0) {
       source = arr[0];
@@ -15432,6 +15482,7 @@ var LumaUni1Max = "luma-uni-1-max";
 var Lyria3Clip = "lyria-3-clip";
 var Lyria3Pro = "lyria-3-pro";
 var Lyria35 = "lyria-3.5";
+var MacaronCookie = "macaron-cookie";
 var MinimaxH3 = "minimax-h3";
 var MinimaxH3ContextIr = "minimax-h3-context-ir";
 var MinimaxH3Max = "minimax-h3-max";
@@ -15706,6 +15757,7 @@ var Models = {
   Lyria3Clip,
   Lyria3Pro,
   Lyria35,
+  MacaronCookie,
   MinimaxH3,
   MinimaxH3ContextIr,
   MinimaxH3Max,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@picsart/ai-sdk",
-  "version": "6.48.1",
+  "version": "6.49.0",
   "type": "module",
   "description": "Type-safe SDK for 100+ AI models — image, video, audio, and text generation with Picsart",
   "license": "MIT",