@picsart/ai-sdk 6.11.0 → 6.12.0

package/index.d.ts

@@ -204,6 +204,16 @@ type ModelInputById = {
         styleExaggeration?: number;
         speed?: number;
         useSpeakerBoost?: boolean;
+        withTimestamps?: boolean;
+    };
+    "eleven-speech-to-text": {
+        audioUrl: string;
+        language?: string;
+        diarize?: boolean;
+        numSpeakers?: number;
+        timestampsGranularity?: "word" | "character";
+        tagAudioEvents?: boolean;
+        seed?: number;
     };
     "eleven-sts-v2": {
         audioUrl: string;
@@ -232,6 +242,11 @@ type ModelInputById = {
         styleExaggeration?: number;
         speed?: number;
         useSpeakerBoost?: boolean;
+        withTimestamps?: boolean;
+    };
+    "eleven-video-to-music": {
+        videoUrls: [string, ...string[]];
+        prompt?: string;
     };
     "eleven-voice-create": {
         prompt: string;
@@ -1885,9 +1900,36 @@ type ModelInputById = {
 type TypedModelId = keyof ModelInputById;
 type ModelInput<M extends TypedModelId> = ModelInputById[M];
 /** IDs of text-generation (LLM) models — narrows generateText(). */
-type TextModelId = "claude-fable-5" | "claude-fable-5-1" | "claude-haiku-4-5" | "claude-opus-4-8" | "claude-opus-5" | "claude-sonnet-4-5" | "claude-sonnet-4-6" | "claude-sonnet-5" | "gemini-2.5-flash" | "gemini-3-pro" | "gemini-3.5-flash-lite" | "gemini-3.6-flash" | "gemini-3.7-flash" | "gemini-3.8-flash" | "gpt-4.1-mini" | "gpt-4.1-nano" | "gpt-4o" | "gpt-4o-mini" | "gpt-5" | "gpt-5-mini" | "gpt-5.1" | "gpt-5.2" | "gpt-5.5" | "gpt-5.6-luna" | "gpt-5.6-sol" | "gpt-5.6-terra" | "gpt-6-astra";
+type TextModelId = "claude-fable-5" | "claude-fable-5-1" | "claude-haiku-4-5" | "claude-opus-4-8" | "claude-opus-5" | "claude-sonnet-4-5" | "claude-sonnet-4-6" | "claude-sonnet-5" | "eleven-speech-to-text" | "gemini-2.5-flash" | "gemini-3-pro" | "gemini-3.5-flash-lite" | "gemini-3.6-flash" | "gemini-3.7-flash" | "gemini-3.8-flash" | "gpt-4.1-mini" | "gpt-4.1-nano" | "gpt-4o" | "gpt-4o-mini" | "gpt-5" | "gpt-5-mini" | "gpt-5.1" | "gpt-5.2" | "gpt-5.5" | "gpt-5.6-luna" | "gpt-5.6-sol" | "gpt-5.6-terra" | "gpt-6-astra";
 type TextModelInputById = Pick<ModelInputById, TextModelId>;
 
+/**
+ * Per-item vendor metadata promoted from the response. Every key is
+ * best-effort: present when the vendor reports it, absent otherwise.
+ */
+interface GenerateResultItemMetadata {
+    /** Explore image id (recraft explore models) — pass back as `sourceImageId`
+     *  to iterate on this image. */
+    exploreImageId?: string;
+    /** Voice preview id (ElevenLabs voice design/remix) — pass to the vendor's
+     *  create-voice-from-preview step to persist the voice. */
+    generatedVoiceId?: string;
+    /** URL of the generated video's last frame, when the model was asked for it
+     *  (`returnLastFrame`, seedance) — the seed for frame-chaining flows. */
+    lastFrameUrl?: string;
+    /** The draft a Seedance 2.5 Draft generation produced — pass it back
+     *  unchanged as `draftTask` to the matching `-draft-final` model to render
+     *  the final 1080p video (valid 7 days). Wire field names on purpose: the
+     *  worker signs it, so any change is rejected. */
+    draftTask?: SeedanceDraftTask;
+}
+/** A Seedance 2.5 draft reference, as the worker issues it. */
+interface SeedanceDraftTask {
+    id: string;
+    video_input: boolean;
+    signature: string;
+}
+
 /**
  * Drive client — manages folders and file saving in Picsart Drive.
  *
@@ -1937,7 +1979,14 @@ interface DriveAttributes {
     appId?: string;
     appType?: AppType;
 }
-interface GenerationFile {
+/** A saved Seedance 2.5 draft — pass `draftTask` unchanged as the final's
+ *  `draftTask` param until `draftExpiresAt`. */
+interface DriveDraftInfo {
+    draftTask?: SeedanceDraftTask;
+    /** ISO timestamp after which the vendor no longer renders a final from it. */
+    draftExpiresAt?: string;
+}
+interface GenerationFile extends DriveDraftInfo {
     appId?: string;
     appType?: AppType;
     model?: string;
@@ -1952,7 +2001,7 @@ interface DriveMediaItem {
     previewUrl?: string;
     timestamp: number;
 }
-interface DriveFileDetails extends DriveMediaItem {
+interface DriveFileDetails extends DriveMediaItem, DriveDraftInfo {
     createdAt?: string;
     model?: string;
     prompt?: string;
@@ -2389,7 +2438,7 @@ type Provider = AppProvider;
 /** App generation modes. */
 type GenerationMode = 'video' | 'image' | 'audio' | 'text';
 /** App input types. */
-type InputType = 't2v' | 'i2v' | 'v2v' | 'a2v' | 't2i' | 'i2i' | 't2a' | 'v2a' | 'tts' | 'sts' | 'sfx' | 'music' | 't2t' | 'i2t' | 'v2t';
+type InputType = 't2v' | 'i2v' | 'v2v' | 'a2v' | 't2i' | 'i2i' | 't2a' | 'v2a' | 'tts' | 'sts' | 'sfx' | 'music' | 't2t' | 'i2t' | 'v2t' | 'a2t';
 interface ModelFeature {
     label: string;
     variant: 'frame' | 'resolution' | 'audio' | 'duration' | 'input' | 'quality' | 'style' | 'characteristic';
@@ -2709,33 +2758,6 @@ interface CatalogsOptions {
     preload?: boolean;
 }
 
-/**
- * Per-item vendor metadata promoted from the response. Every key is
- * best-effort: present when the vendor reports it, absent otherwise.
- */
-interface GenerateResultItemMetadata {
-    /** Explore image id (recraft explore models) — pass back as `sourceImageId`
-     *  to iterate on this image. */
-    exploreImageId?: string;
-    /** Voice preview id (ElevenLabs voice design/remix) — pass to the vendor's
-     *  create-voice-from-preview step to persist the voice. */
-    generatedVoiceId?: string;
-    /** URL of the generated video's last frame, when the model was asked for it
-     *  (`returnLastFrame`, seedance) — the seed for frame-chaining flows. */
-    lastFrameUrl?: string;
-    /** The draft a Seedance 2.5 Draft generation produced — pass it back
-     *  unchanged as `draftTask` to the matching `-draft-final` model to render
-     *  the final 1080p video (valid 7 days). Wire field names on purpose: the
-     *  worker signs it, so any change is rejected. */
-    draftTask?: SeedanceDraftTask;
-}
-/** A Seedance 2.5 draft reference, as the worker issues it. */
-interface SeedanceDraftTask {
-    id: string;
-    video_input: boolean;
-    signature: string;
-}
-
 /**
  * Failure codes the SDK synthesizes when the platform supplies no `reason`.
  * An API-supplied `reason` passes through unchanged, so the open `string`
@@ -3076,9 +3098,11 @@ declare const Models: {
     readonly ElevenDubbing: "eleven-dubbing";
     readonly ElevenMultilingualStsV2: "eleven-multilingual-sts-v2";
     readonly ElevenMultilingualV2: "eleven-multilingual-v2";
+    readonly ElevenSpeechToText: "eleven-speech-to-text";
     readonly ElevenStsV2: "eleven-sts-v2";
     readonly ElevenTextToDialogue: "eleven-text-to-dialogue";
     readonly ElevenV3: "eleven-v3";
+    readonly ElevenVideoToMusic: "eleven-video-to-music";
     readonly ElevenVoiceCreate: "eleven-voice-create";
     readonly ElevenVoiceDesignV2: "eleven-voice-design-v2";
     readonly ElevenVoiceDesignV3: "eleven-voice-design-v3";
@@ -3443,4 +3467,4 @@ declare const getModel: (id: string) => ModelDefinition | undefined;
  */
 declare const findModel: (ref: string) => ModelDefinition | undefined;
 
-export { ALL_MODELS, type AiClient, ApiError, type ApiErrorCode, type ApiErrorInit, type ApiResponse, type ApiRunOptions, type ApiSchemas, type ApisClient, type AppIdentity, type AppType, type AuthenticatedFetch, type AvatarOption, type BooleanDescriptor, type BooleanEntry, type CatalogDescriptor, type CatalogEntry, type CatalogItem, type CatalogPage, type CatalogPageOptions, type CatalogPreview, type CatalogQuery, type CatalogResult, type CatalogSource, type CatalogsClient, type CatalogsOptions, type ClientConfig, type CreditRange, type CreditRangeContext, type CreditTier, type CreditUsage, DEFAULT_VISIBLE_RELEASES, type DeepLinkResult, type DriveAttributes, type DriveClient, type DriveConfig, type DriveFile, type DriveFileDetails, type DriveFolder, type DriveMediaItem, type DriveSaveResult, type EntryMeta, type EnumDescriptor, type EnumEntry, type EnumOption, type FileDescriptor, type FileEntry, type FlatParamEntry, type GenerateOptions, type GenerateResult, type GenerateResultItem, type GenerateResultItemMetadata, type GenerateTextResult, type GenerationContext, type GenerationEvent, GenerationEventType, type GenerationFile, type GenerationMode, type GenerationOptions, type GenerationProgress, type ListOptions, type MediaModelId, type MediaTypeFilter, Model, type ModelDefinition, type ModelDescriptor, type ModelFilter, type ModelInput, type ModelInputById, type ModelMeta, type ModelParams, type ModelParamsAccessor, Models, type ObjectDescriptor, type ObjectEntry, type ParamDescriptor, type ParamEntry, type ParamOption, type PayloadDriveFolderOptions, type PayloadDriveOptions, type PayloadInputsTransformationOptions, type PricingOptions, type ProviderInfo, type RangeDescriptor, type RangeEntry, type ReleaseTag, type SaveParams, type SdkPayload, type SdkTransport, type SeedanceDraftTask, type TextDescriptor, type TextEntry, type TextModelId, type TextModelInputById, type ToolUsage, type TransportPollOptions, type TransportResult, type TypedModelId, type UserReaction, type ValidationResult, type VoiceOption, type WorkflowJobHandle, type WorkflowSubmitRequest, buildFilename, buildGenerationAttributes, catalog, createClient, decodeDeepLinkPayload, encodeDeepLinkPayload, findModel, getModel, getModelsByMode, getVoiceById, inferResourceType, isVisibleForReleases, parseGeneration, releaseOf, toAvatarOption, toVoiceOption };
+export { ALL_MODELS, type AiClient, ApiError, type ApiErrorCode, type ApiErrorInit, type ApiResponse, type ApiRunOptions, type ApiSchemas, type ApisClient, type AppIdentity, type AppType, type AuthenticatedFetch, type AvatarOption, type BooleanDescriptor, type BooleanEntry, type CatalogDescriptor, type CatalogEntry, type CatalogItem, type CatalogPage, type CatalogPageOptions, type CatalogPreview, type CatalogQuery, type CatalogResult, type CatalogSource, type CatalogsClient, type CatalogsOptions, type ClientConfig, type CreditRange, type CreditRangeContext, type CreditTier, type CreditUsage, DEFAULT_VISIBLE_RELEASES, type DeepLinkResult, type DriveAttributes, type DriveClient, type DriveConfig, type DriveDraftInfo, type DriveFile, type DriveFileDetails, type DriveFolder, type DriveMediaItem, type DriveSaveResult, type EntryMeta, type EnumDescriptor, type EnumEntry, type EnumOption, type FileDescriptor, type FileEntry, type FlatParamEntry, type GenerateOptions, type GenerateResult, type GenerateResultItem, type GenerateResultItemMetadata, type GenerateTextResult, type GenerationContext, type GenerationEvent, GenerationEventType, type GenerationFile, type GenerationMode, type GenerationOptions, type GenerationProgress, type ListOptions, type MediaModelId, type MediaTypeFilter, Model, type ModelDefinition, type ModelDescriptor, type ModelFilter, type ModelInput, type ModelInputById, type ModelMeta, type ModelParams, type ModelParamsAccessor, Models, type ObjectDescriptor, type ObjectEntry, type ParamDescriptor, type ParamEntry, type ParamOption, type PayloadDriveFolderOptions, type PayloadDriveOptions, type PayloadInputsTransformationOptions, type PricingOptions, type ProviderInfo, type RangeDescriptor, type RangeEntry, type ReleaseTag, type SaveParams, type SdkPayload, type SdkTransport, type SeedanceDraftTask, type TextDescriptor, type TextEntry, type TextModelId, type TextModelInputById, type ToolUsage, type TransportPollOptions, type TransportResult, type TypedModelId, type UserReaction, type ValidationResult, type VoiceOption, type WorkflowJobHandle, type WorkflowSubmitRequest, buildFilename, buildGenerationAttributes, catalog, createClient, decodeDeepLinkPayload, encodeDeepLinkPayload, findModel, getModel, getModelsByMode, getVoiceById, inferResourceType, isVisibleForReleases, parseGeneration, releaseOf, toAvatarOption, toVoiceOption };

package/index.js

@@ -6937,7 +6937,11 @@ var ttsParamConfig = (promptMaxLength, withLanguage) => ({
   ...withLanguage ? params.language(false) : {},
   ...params.prompt({ maxLength: promptMaxLength }),
   ...params.voiceId([], DEFAULT_VOICE_ID, { catalog: VOICE_CATALOG }),
-  ...voiceSettingsParams(true)
+  ...voiceSettingsParams(true),
+  // Timings come back beside the audio, for captions and lip sync. Off by
+  // default: the vendor answers a different route that inlines the audio as
+  // base64, and callers who do not need timings should not pay the decode.
+  ...p.boolean("withTimestamps", false, "Character Timings")
 });
 var { MODELS: MODELS24 } = defineModels("elevenlabs", [
   // ── TTS ───────────────────────────────────────────────────────────
@@ -7063,6 +7067,56 @@ var { MODELS: MODELS24 } = defineModels("elevenlabs", [
       ...p.boolean("isInstrumental", false, "Instrumental Only")
     }
   },
+  // ── Transcription ─────────────────────────────────────────────────
+  {
+    // `modelId` is the pricing-catalog key: the vendor id `scribe_v2` is not
+    // vendor-scoped, so pricing carries the `eleven_` prefix, as music does.
+    id: "eleven-speech-to-text",
+    name: "Eleven Scribe v2",
+    modelId: "eleven_scribe_v2",
+    addedAt: "2026-09-24",
+    workflow: "elevenlabs/v1/speech-to-text",
+    estimatedTime: 10,
+    mode: "text",
+    inputType: "a2t",
+    description: "Transcribe speech from audio or video, with word timings and speaker labels.",
+    features: [feat("Transcription", "characteristic"), feat("Speaker Labels", "characteristic")],
+    paramConfig: {
+      ...params.audioInput("Audio or Video", true),
+      // ISO 639-1 or 639-3. Left free text: the vendor detects the language on
+      // its own, and the closed list would be 99 entries of upkeep.
+      language: {
+        label: "Language (ISO code, optional)",
+        descriptor: { kind: "text", placeholder: "e.g. en, rus \u2014 omit to auto-detect" }
+      },
+      ...p.boolean("diarize", false, "Label Speakers"),
+      ...p.range("numSpeakers", 1, 32, 1, { step: 1, label: "Speakers" }),
+      ...p.enum("timestampsGranularity", ["word", "character"], "word", { label: "Timing Detail" }),
+      ...p.boolean("tagAudioEvents", false, "Tag Audio Events"),
+      ...params.seed(2147483647)
+    }
+  },
+  // ── Video to Music ────────────────────────────────────────────────
+  {
+    // Same pricing model as music-generation — the same engine writes the
+    // track; what separates the two is the use case, `video-to-audio`.
+    id: "eleven-video-to-music",
+    name: "Eleven Video to Music",
+    modelId: "eleven_music_v2",
+    addedAt: "2026-09-24",
+    workflow: "elevenlabs/v1/video-to-music",
+    estimatedTime: 25,
+    mode: "audio",
+    inputType: "v2a",
+    description: "Score a video with a soundtrack written to follow what happens on screen.",
+    features: [feat("Video Scoring", "characteristic"), feat("Soundtrack", "characteristic")],
+    paramConfig: {
+      // The vendor takes up to ten clips, 200MB and 600 seconds in total, and
+      // scores them as one timeline.
+      ...params.videoInputs(10, "Videos", true),
+      ...params.prompt({ maxLength: 1e3, required: false, placeholder: "How the soundtrack should sound" })
+    }
+  },
   // ── Speech-to-Speech ──────────────────────────────────────────────
   {
     id: "eleven-sts-v2",
@@ -7227,6 +7281,7 @@ var buildElevenLabsTTSPayload = (modelId) => (input) => ({
   voice_id: input.voiceId ?? DEFAULT_VOICE_ID,
   model_id: modelId,
   ...input.language ? { language_code: input.language } : {},
+  ...input.withTimestamps ? { with_timestamps: true } : {},
   ...voiceSettings(input)
 });
 var buildElevenLabsSTSPayload = (modelId) => (input) => ({
@@ -7246,13 +7301,30 @@ var buildElevenLabsDialoguePayload = (input) => ({
   ...input.language ? { language_code: input.language } : {},
   ...input.seed != null ? { seed: input.seed } : {}
 });
+var buildElevenLabsTranscribePayload = (input) => ({
+  audio_url: input.audioUrl,
+  ...input.language ? { language_code: input.language } : {},
+  ...input.diarize ? { diarize: true } : {},
+  // Only meaningful alongside diarization, and only when the caller moved it
+  // off the default of one.
+  ...input.diarize && input.numSpeakers && input.numSpeakers > 1 ? { num_speakers: input.numSpeakers } : {},
+  ...input.timestampsGranularity ? { timestamps_granularity: input.timestampsGranularity } : {},
+  ...input.tagAudioEvents ? { tag_audio_events: true } : {},
+  ...input.seed != null ? { seed: input.seed } : {}
+});
+var buildElevenLabsVideoToMusicPayload = (input) => ({
+  video_urls: input.videoUrls,
+  ...input.prompt ? { description: input.prompt } : {}
+});
 registerPayloads(MODELS24, {
   "elevenlabs-music-v2": buildElevenLabsMusicPayload,
   "eleven-v3": buildElevenLabsTTSPayload("eleven_v3"),
   "eleven-multilingual-v2": buildElevenLabsTTSPayload("eleven_multilingual_v2"),
   "eleven-sts-v2": buildElevenLabsSTSPayload("eleven_english_sts_v2"),
   "eleven-multilingual-sts-v2": buildElevenLabsSTSPayload("eleven_multilingual_sts_v2"),
-  "eleven-text-to-dialogue": buildElevenLabsDialoguePayload
+  "eleven-text-to-dialogue": buildElevenLabsDialoguePayload,
+  "eleven-speech-to-text": buildElevenLabsTranscribePayload,
+  "eleven-video-to-music": buildElevenLabsVideoToMusicPayload
 });
 
 // src/vendors/catalog/heygen.ts
@@ -11588,6 +11660,8 @@ var MAX_DRIVE_PROMPT_LENGTH = 18e3;
 
 // src/client/drive.ts
 var USER_REACTION_ATTR = "userReaction";
+var DRAFT_TASK_ATTR = "draft_task";
+var DRAFT_EXPIRES_AT_ATTR = "draft_expires_at";
 function inferResourceType(mode) {
   if (mode === "video") return "VIDEO";
   if (mode === "audio") return "AUDIO";
@@ -11634,6 +11708,14 @@ function parseAttributes(raw) {
   }
   return map;
 }
+function parseDraftInfo(attrs) {
+  const draftTask = parseJsonAttr(attrs[DRAFT_TASK_ATTR]);
+  if (!isSeedanceDraftTask(draftTask)) return {};
+  return {
+    draftTask: { id: draftTask.id, video_input: draftTask.video_input, signature: draftTask.signature },
+    ...attrs[DRAFT_EXPIRES_AT_ATTR] ? { draftExpiresAt: attrs[DRAFT_EXPIRES_AT_ATTR] } : {}
+  };
+}
 function parseReaction(value) {
   return value === "like" || value === "dislike" ? value : void 0;
 }
@@ -11706,7 +11788,8 @@ function toDetailedItem(file) {
     referenceAudioUrl: extras.referenceAudioUrl,
     aspectRatio: extras.aspectRatio,
     resolution: extras.resolution,
-    quality: extras.quality
+    quality: extras.quality,
+    ...parseDraftInfo(attrs)
   };
 }
 var LEGACY_TOOL_APP = {
@@ -11751,7 +11834,8 @@ function adaptLegacyGeneration(attrs) {
     appType: app?.appType,
     model: attrs.model || void 0,
     aiSDKPayload,
-    userReaction: parseReaction(attrs[USER_REACTION_ATTR])
+    userReaction: parseReaction(attrs[USER_REACTION_ATTR]),
+    ...parseDraftInfo(attrs)
   };
 }
 function parseGeneration(file) {
@@ -11764,7 +11848,8 @@ function parseGeneration(file) {
     appType: attrs.appType === "native" || attrs.appType === "miniapp" ? attrs.appType : void 0,
     model: attrs.model || void 0,
     aiSDKPayload: parseJsonAttr(attrs.aiSDKPayload),
-    userReaction: parseReaction(attrs[USER_REACTION_ATTR])
+    userReaction: parseReaction(attrs[USER_REACTION_ATTR]),
+    ...parseDraftInfo(attrs)
   };
 }
 function createDriveClient(f, apiUrl, rootFolderName) {
@@ -12574,9 +12659,11 @@ var ElevenAudioIsolation = "eleven-audio-isolation";
 var ElevenDubbing = "eleven-dubbing";
 var ElevenMultilingualStsV2 = "eleven-multilingual-sts-v2";
 var ElevenMultilingualV2 = "eleven-multilingual-v2";
+var ElevenSpeechToText = "eleven-speech-to-text";
 var ElevenStsV2 = "eleven-sts-v2";
 var ElevenTextToDialogue = "eleven-text-to-dialogue";
 var ElevenV3 = "eleven-v3";
+var ElevenVideoToMusic = "eleven-video-to-music";
 var ElevenVoiceCreate = "eleven-voice-create";
 var ElevenVoiceDesignV2 = "eleven-voice-design-v2";
 var ElevenVoiceDesignV3 = "eleven-voice-design-v3";
@@ -12819,9 +12906,11 @@ var Models = {
   ElevenDubbing,
   ElevenMultilingualStsV2,
   ElevenMultilingualV2,
+  ElevenSpeechToText,
   ElevenStsV2,
   ElevenTextToDialogue,
   ElevenV3,
+  ElevenVideoToMusic,
   ElevenVoiceCreate,
   ElevenVoiceDesignV2,
   ElevenVoiceDesignV3,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@picsart/ai-sdk",
-  "version": "6.11.0",
+  "version": "6.12.0",
   "type": "module",
   "description": "Type-safe SDK for 100+ AI models — image, video, audio, and text generation with Picsart",
   "license": "MIT",