---
title: experimental_streamTranslate
description: API Reference for experimental_streamTranslate.
url: "https://ai-sdk.dev/docs/reference/ai-sdk-core/stream-translate"
docs_index: /llms.txt
---

> For an index of all documentation, see [/llms.txt](/llms.txt).

`experimental_streamTranslate` is an experimental feature.

Streams a speech-to-speech translation from live raw audio. Models translate
live source audio into target-language audio and text.

`experimental_streamTranslate` is built on the speech translation model
specification (`Experimental_SpeechTranslationModelV4`). Provider
implementations of the specification ship separately — see your provider's
documentation for available translation models.

```ts
import { experimental_streamTranslate as streamTranslate } from 'ai';

const result = streamTranslate({
  // any provider model instance that implements the experimental
  // speech translation model specification (Experimental_SpeechTranslationModelV4):
  model: translationModel,
  audio: audioStream, // ReadableStream<Uint8Array | string>
  inputAudioFormat: { type: 'audio/pcm', rate: 24000 },
  targetLanguage: 'es',
});

for await (const part of result.fullStream) {
  if (part.type === 'output-text-delta') {
    process.stdout.write(part.delta);
  }
}

console.log(await result.translationText);
```

## Import

```
import { experimental_streamTranslate as streamTranslate } from "ai"
```

## API Signature

### Parameters

- `model` (`Experimental_SpeechTranslationModelV4`): The speech translation model to use. Translation is a streaming-only modality. Pass a provider model instance that implements the experimental speech translation model specification. String model IDs resolve through the global provider (AI Gateway by default) when it supports speech translation models.
- `audio` (`ReadableStream<Uint8Array | string>`): Raw audio chunks to translate. \`Uint8Array\` chunks contain raw audio bytes; \`string\` chunks contain base64-encoded raw audio bytes.
- `inputAudioFormat` (`{ type: string; rate?: number }`): The input audio format for the raw audio chunks, e.g. \`\{ type: "audio/pcm", rate: 24000 }\`. Supported types are provider-specific (e.g. \`audio/pcm\`, \`audio/pcmu\`, \`audio/pcma\`).
- `targetLanguage` (`string`): The language to translate the audio into, as a BCP-47-style language tag (e.g. \`en\`, \`es\`, \`fr-CA\`). Supported values are provider-specific and validated by the provider.
- `sourceLanguage?` (`string`): The language of the source audio, as a BCP-47-style language tag. When absent, providers auto-detect the source language.
- `outputAudioFormat?` (`{ type: string; rate?: number }`): The desired audio format for translated audio chunks. When absent, the provider default output format is used.
- `providerOptions?` (`Record<string, JSONObject>`): Additional provider-specific options.
- `abortSignal?` (`AbortSignal`): An optional abort signal to cancel the call.
- `headers?` (`Record<string, string>`): Additional HTTP/WebSocket headers, if supported by the provider.
- `includeRawChunks?` (`boolean`): When true, the provider includes raw provider chunks in the stream as \`raw\` parts.

### Returns

- `fullStream` (`AsyncIterableStream<TranslationStreamPart>`): A single-consumer live stream of translation parts: \`audio\`, \`output-text-delta\`, \`output-text-final\`, \`source-transcript-delta\`, \`source-transcript-partial\`, \`source-transcript-final\`, \`raw\`, and \`error\`. Access it once, before any result promise, when both stream parts and final results are needed. Accessing a result promise first consumes the stream internally and makes \`fullStream\` unavailable.
- `sourceText` (`Promise<string>`): The final source-language transcript of the input audio.
- `translationText` (`Promise<string>`): The final translated text in the target language. May resolve to an empty string for providers that produce only audio output.
- `durationInSeconds` (`Promise<number | undefined>`): The duration of the source audio in seconds, if available.
- `usage` (`Promise<Experimental_SpeechTranslationModelV4Usage | undefined>`): Usage information for the translation call, if reported by the provider: \`inputAudioSeconds\`, \`inputAudioTokens\`, \`outputAudioTokens\`, \`inputTextTokens\`, \`outputTextTokens\` (all optional numbers).
- `warnings` (`Promise<Warning[]>`): Warnings for the call, e.g. unsupported settings. Resolves at stream start, or with an empty array at finish if no stream-start was emitted.
- `response` (`Promise<SpeechTranslationModelResponseMetadata>`): Response metadata (timestamp, model ID, headers).
- `providerMetadata` (`Promise<Record<string, JSONObject>>`): Additional provider-specific metadata.

The result promises settle as the stream is consumed. If you stop consuming
`fullStream` early (e.g. `break` out of the loop), the underlying provider
connection is closed and pending result promises reject.

---

For a semantic overview of all documentation, see [/sitemap.md](/sitemap.md)

For an index of all available documentation, see [/llms.txt](/llms.txt)

For agent-facing discovery, including API and MCP surfaces, see [/agents.md](/agents.md)