@ai-sdk/google 4.0.53 → 4.0.54

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ai-sdk/google",
-  "version": "4.0.53",
+  "version": "4.0.54",
   "type": "module",
   "license": "Apache-2.0",
   "sideEffects": false,

package/src/google-provider.ts

@@ -10,6 +10,7 @@ import type {
   Experimental_RealtimeFactoryV4GetTokenOptions as RealtimeFactoryV4GetTokenOptions,
   SpeechModelV4,
   Experimental_SpeechTranslationModelV4 as SpeechTranslationModelV4,
+  TranscriptionModelV4,
 } from '@ai-sdk/provider';
 import {
   generateId,
@@ -43,6 +44,8 @@ import {
 import type { GoogleInteractionsModelId } from './interactions/google-interactions-language-model-options';
 import type { GoogleInteractionsAgentName } from './interactions/google-interactions-agent';
 import { GoogleRealtimeModel } from './realtime/google-realtime-model';
+import { GoogleTranscriptionModel } from './transcription/google-transcription-model';
+import type { GoogleTranscriptionModelId } from './transcription/google-transcription-model-options';
 import { GoogleSpeechTranslationModel } from './speech-translation/google-speech-translation-model';
 import type { GoogleSpeechTranslationModelId } from './speech-translation/google-speech-translation-model-options';
 
@@ -120,6 +123,19 @@ export interface GoogleProvider extends ProviderV4 {
    */
   speechModel(modelId: GoogleSpeechModelId): SpeechModelV4;
 
+  /**
+   * Creates a model for transcription (speech-to-text). Unary models
+   * (e.g. `gemini-3.5-transcribe`) transcribe audio files; live models
+   * (e.g. `gemini-3.5-transcribe-live`) stream transcription over the
+   * Gemini Live API WebSocket via `experimental_streamTranscribe`.
+   */
+  transcription(modelId: GoogleTranscriptionModelId): TranscriptionModelV4;
+
+  /**
+   * Creates a model for transcription (speech-to-text).
+   */
+  transcriptionModel(modelId: GoogleTranscriptionModelId): TranscriptionModelV4;
+
   files(): FilesV4;
 
   /**
@@ -331,6 +347,15 @@ export function createGoogle(
       fetch: options.fetch,
     });
 
+  const createTranscriptionModel = (modelId: GoogleTranscriptionModelId) =>
+    new GoogleTranscriptionModel(modelId, {
+      provider: `${providerName}.transcription`,
+      baseURL,
+      headers: getHeaders,
+      fetch: options.fetch,
+      webSocket: options.webSocket,
+    });
+
   const experimentalRealtimeFactory = Object.assign(
     (modelId: string) => createRealtimeModel(modelId),
     {
@@ -393,6 +418,8 @@ export function createGoogle(
   provider.files = createFiles;
   provider.speech = createSpeechModel;
   provider.speechModel = createSpeechModel;
+  provider.transcription = createTranscriptionModel;
+  provider.transcriptionModel = createTranscriptionModel;
   provider.translation = createSpeechTranslationModel;
   provider.speechTranslationModel = createSpeechTranslationModel;
   provider.interactions = createInteractionsModel;

package/src/index.ts

@@ -60,6 +60,11 @@ export type {
   GoogleRealtimeModelId as Experimental_GoogleRealtimeModelId,
   GoogleRealtimeModelOptions as Experimental_GoogleRealtimeModelOptions,
 } from './realtime/google-realtime-model-options';
+export { GoogleTranscriptionModel } from './transcription/google-transcription-model';
+export type {
+  GoogleTranscriptionModelId,
+  GoogleTranscriptionModelOptions,
+} from './transcription/google-transcription-model-options';
 export {
   GoogleSpeechTranslationModel as Experimental_GoogleSpeechTranslationModel,
   /** @deprecated Use `Experimental_GoogleSpeechTranslationModel` instead. */

package/src/transcription/google-transcription-model-options.ts

@@ -0,0 +1,50 @@
+import { z } from 'zod/v4';
+
+export type GoogleTranscriptionModelId =
+  | 'gemini-3.5-transcribe'
+  | 'gemini-3.5-transcribe-live'
+  | (string & {});
+
+/**
+ * Speech recognition options shared by unary (`gemini-3.5-transcribe`) and
+ * live (`gemini-3.5-transcribe-live`) transcription. Maps onto Google's
+ * `AudioTranscriptionConfig`.
+ */
+export const googleTranscriptionModelOptions = z.object({
+  /**
+   * BCP-47 language codes providing hints about the languages present in the
+   * audio. If omitted or empty, defaults to automatic language detection.
+   */
+  languageCodes: z.array(z.string()).optional(),
+
+  /**
+   * Custom vocabulary phrases, which bias the speech recognition model
+   * toward recognizing specific terms.
+   */
+  customVocabulary: z.array(z.string()).optional(),
+
+  /**
+   * Enables word-level timestamp generation.
+   */
+  wordTimestamp: z.boolean().optional(),
+
+  /**
+   * Enables speaker diarization.
+   */
+  diarization: z.boolean().optional(),
+
+  /**
+   * Transcription output formatting mode.
+   *
+   * - `VERBATIM` (default): exact literal transcript preserving filler
+   *   words, repetitions, and false starts.
+   * - `SMART`: cleans up and structures the transcript in real time —
+   *   disfluency removal, inline self-corrections, structured formatting
+   *   (lists, numbers, dates, paragraph breaks), and grammar/casing polish.
+   */
+  mode: z.enum(['SMART', 'VERBATIM']).optional(),
+});
+
+export type GoogleTranscriptionModelOptions = z.infer<
+  typeof googleTranscriptionModelOptions
+>;