@picsart/ai-sdk 5.18.0 → 5.19.0

package/_vendor/workflows-types/index.d.ts

@@ -109,7 +109,7 @@ interface SpeakerVoiceConfig$2 {
     speaker: string;
     voiceConfig: VoiceConfig$2;
 }
-type GeminiModel$1 = "gemini-2.5-pro" | "gemini-2.5-flash" | "gemini-2.5-flash-image-preview" | "gemini-2.5-flash-image" | "gemini-2.5-flash-lite-preview-06-17" | "gemini-2.0-flash" | "gemini-2.0-flash-preview-image-generation" | "gemini-2.0-flash-lite" | "gemini-1.5-flash" | "gemini-1.5-flash-8b" | "gemini-1.5-pro" | "gemini-embedding-exp" | "gemini-3-pro" | "gemini-3-pro-preview" | "gemini-3.1-pro-preview" | "gemini-3.1-flash-lite" | "gemini-3-pro-image" | "gemini-3-pro-image-preview" | "gemini-3.1-flash-image-preview" | "gemini-3.1-flash-image" | "gemini-3.1-flash-lite-image" | "gemini-2.5-flash-tts" | "gemini-2.5-pro-tts" | "instant-ramen";
+type GeminiModel$1 = "gemini-2.5-pro" | "gemini-2.5-flash" | "gemini-2.5-flash-image-preview" | "gemini-2.5-flash-image" | "gemini-2.5-flash-lite-preview-06-17" | "gemini-2.0-flash" | "gemini-2.0-flash-preview-image-generation" | "gemini-2.0-flash-lite" | "gemini-1.5-flash" | "gemini-1.5-flash-8b" | "gemini-1.5-pro" | "gemini-embedding-exp" | "gemini-3-pro" | "gemini-3-pro-preview" | "gemini-3.1-pro-preview" | "gemini-3.1-flash-lite" | "gemini-3.1-flash-lite-preview" | "gemini-3-pro-image" | "gemini-3-pro-image-preview" | "gemini-3.1-flash-image-preview" | "gemini-3.1-flash-image" | "gemini-3.1-flash-lite-image" | "gemini-2.5-flash-tts" | "gemini-2.5-pro-tts" | "instant-ramen";
 interface GeminiResult$1 {
     id: string;
     status: "ACCEPTED" | "IN_PROGRESS" | "COMPLETED" | "FAILED";
@@ -435,7 +435,7 @@ interface SpeakerVoiceConfig$1 {
     speaker: string;
     voiceConfig: VoiceConfig$1;
 }
-type GeminiModel = "gemini-2.5-pro" | "gemini-2.5-flash" | "gemini-2.5-flash-image-preview" | "gemini-2.5-flash-image" | "gemini-2.5-flash-lite-preview-06-17" | "gemini-2.0-flash" | "gemini-2.0-flash-preview-image-generation" | "gemini-2.0-flash-lite" | "gemini-1.5-flash" | "gemini-1.5-flash-8b" | "gemini-1.5-pro" | "gemini-embedding-exp" | "gemini-3-pro" | "gemini-3-pro-preview" | "gemini-3.1-pro-preview" | "gemini-3.1-flash-lite" | "gemini-3-pro-image" | "gemini-3-pro-image-preview" | "gemini-3.1-flash-image-preview" | "gemini-3.1-flash-image" | "gemini-3.1-flash-lite-image" | "gemini-2.5-flash-tts" | "gemini-2.5-pro-tts" | "instant-ramen";
+type GeminiModel = "gemini-2.5-pro" | "gemini-2.5-flash" | "gemini-2.5-flash-image-preview" | "gemini-2.5-flash-image" | "gemini-2.5-flash-lite-preview-06-17" | "gemini-2.0-flash" | "gemini-2.0-flash-preview-image-generation" | "gemini-2.0-flash-lite" | "gemini-1.5-flash" | "gemini-1.5-flash-8b" | "gemini-1.5-pro" | "gemini-embedding-exp" | "gemini-3-pro" | "gemini-3-pro-preview" | "gemini-3.1-pro-preview" | "gemini-3.1-flash-lite" | "gemini-3.1-flash-lite-preview" | "gemini-3-pro-image" | "gemini-3-pro-image-preview" | "gemini-3.1-flash-image-preview" | "gemini-3.1-flash-image" | "gemini-3.1-flash-lite-image" | "gemini-2.5-flash-tts" | "gemini-2.5-pro-tts" | "instant-ramen";
 interface GeminiResult {
     id: string;
     status: "ACCEPTED" | "IN_PROGRESS" | "COMPLETED" | "FAILED";
@@ -6762,22 +6762,22 @@ interface HappyhorseImageToVideoResult {
 interface WhisperxVideoCaptionsCommand {
     url: string;
     model?: "preview-model-whisperx-large-v3" | "model-whisperx-large-v3";
-    parameters?: WhisperParameters;
-    alignment?: AlignmentOptions;
-    diarization?: DiarizationOptions;
-    output?: OutputOptions;
+    parameters?: WhisperParameters$1;
+    alignment?: AlignmentOptions$1;
+    diarization?: DiarizationOptions$1;
+    output?: OutputOptions$1;
     options?: GenAIOptions$19;
 }
-interface WhisperParameters {
+interface WhisperParameters$1 {
     language?: "en" | "es" | "fr" | "de" | "it" | "pt" | "nl" | "pl" | "ja" | "zh" | "ru";
     task?: "transcribe" | "translate";
 }
-interface AlignmentOptions {
+interface AlignmentOptions$1 {
     enabled?: boolean;
     return_char_alignments?: boolean;
     interpolate_method?: "nearest" | "linear" | "ignore";
 }
-interface DiarizationOptions {
+interface DiarizationOptions$1 {
     enabled?: boolean;
     num_speakers?: number;
     min_speakers?: number;
@@ -6786,7 +6786,7 @@ interface DiarizationOptions {
     fill_nearest?: boolean;
     apply_word_speakers?: boolean;
 }
-interface OutputOptions {
+interface OutputOptions$1 {
     include_word_segments?: boolean;
     include_char_segments?: boolean;
 }
@@ -6813,24 +6813,24 @@ interface WhisperxVideoCaptionsResponse {
 }
 interface WhisperxVideoCaptionsResultData {
     language: string;
-    segments?: TranscriptionSegment[];
+    segments?: TranscriptionSegment$1[];
 }
-interface TranscriptionSegment {
+interface TranscriptionSegment$1 {
     start: number;
     end: number;
     text: string;
     speaker?: string;
-    words?: WordSegment[];
-    chars?: CharSegment[];
+    words?: WordSegment$1[];
+    chars?: CharSegment$1[];
 }
-interface WordSegment {
+interface WordSegment$1 {
     word: string;
     start?: number;
     end?: number;
     score?: number;
     speaker?: string;
 }
-interface CharSegment {
+interface CharSegment$1 {
     char: string;
     start?: number;
     end?: number;
@@ -9218,7 +9218,7 @@ interface ImageSegmentationResultData {
 interface WanV3VideoCommand {
     prompt?: string;
     media?: WanV3MediaItem[];
-    model?: "wan3.0-video";
+    model?: "wan3.0-video" | "wan3.0-video-prime";
     resolution?: "480P" | "720P" | "1080P";
     ratio?: "16:9" | "9:16" | "1:1" | "4:3" | "3:4" | "adaptive";
     duration?: number;
@@ -10333,6 +10333,69 @@ interface FaceCollageResultData {
     driveFile?: Record<string, unknown>;
 }
 
+interface WhisperxSliceCommand {
+    url: string;
+    start?: number;
+    end?: number;
+    model?: "preview-model-whisperx-large-v3" | "model-whisperx-large-v3";
+    parameters?: WhisperParameters;
+    alignment?: AlignmentOptions;
+    diarization?: DiarizationOptions;
+    output?: OutputOptions;
+}
+interface WhisperParameters {
+    language?: "en" | "es" | "fr" | "de" | "it" | "pt" | "nl" | "pl" | "ja" | "zh" | "ru";
+    task?: "transcribe" | "translate";
+}
+interface AlignmentOptions {
+    enabled?: boolean;
+    return_char_alignments?: boolean;
+    interpolate_method?: "nearest" | "linear" | "ignore";
+}
+interface DiarizationOptions {
+    enabled?: boolean;
+    num_speakers?: number;
+    min_speakers?: number;
+    max_speakers?: number;
+    return_embeddings?: boolean;
+    fill_nearest?: boolean;
+    apply_word_speakers?: boolean;
+}
+interface OutputOptions {
+    include_word_segments?: boolean;
+    include_char_segments?: boolean;
+}
+interface WhisperxSliceResponse {
+    id: string;
+    status: "ACCEPTED" | "IN_PROGRESS" | "COMPLETED" | "FAILED";
+    result: WhisperxSliceResultData;
+}
+interface WhisperxSliceResultData {
+    language: string;
+    segments?: TranscriptionSegment[];
+}
+interface TranscriptionSegment {
+    start: number;
+    end: number;
+    text: string;
+    speaker?: string;
+    words?: WordSegment[];
+    chars?: CharSegment[];
+}
+interface WordSegment {
+    word: string;
+    start?: number;
+    end?: number;
+    score?: number;
+    speaker?: string;
+}
+interface CharSegment {
+    char: string;
+    start?: number;
+    end?: number;
+    score?: number;
+}
+
 interface WorkflowTypes {
     'check-text': {
         params: CheckTextCommand;
@@ -11334,6 +11397,10 @@ interface WorkflowTypes {
         params: FaceCollageCommand;
         result: FaceCollageResponse;
     };
+    'pcp/v1/whisperx-stt-slice': {
+        params: WhisperxSliceCommand;
+        result: WhisperxSliceResponse;
+    };
 }
 
 export type { WorkflowTypes };

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";
@@ -2492,6 +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";

package/index.js

@@ -5753,9 +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",
@@ -10465,6 +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";
@@ -10672,6 +10720,7 @@ var Models = {
   Gemini35FlashLite,
   Gemini36Flash,
   Gemini37Flash,
+  GeminiOmni11FlashPreview,
   GeminiOmniFlashPreview,
   Gpt55,
   GptImage1,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@picsart/ai-sdk",
-  "version": "5.18.0",
+  "version": "5.19.0",
   "type": "module",
   "description": "Type-safe SDK for 100+ AI models — image, video, audio, and text generation with Picsart",
   "license": "MIT",