> ## Documentation Index
> Fetch the complete documentation index at: https://bianxieai.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Create a transcription

> Transcribe audio in its original language

Send `multipart/form-data`.

| Field                     | Type   | Required | Description                                                       |
| ------------------------- | ------ | -------- | ----------------------------------------------------------------- |
| `file`                    | file   | Yes      | Audio file; supported formats and size depend on the model.       |
| `model`                   | string | Yes      | Transcription model ID.                                           |
| `language`                | string | No       | ISO-639-1 input language.                                         |
| `prompt`                  | string | No       | Style guidance or prior context.                                  |
| `response_format`         | string | No       | json, text, srt, verbose\_json, or vtt, subject to model support. |
| `temperature`             | number | No       | Sampling temperature.                                             |
| `timestamp_granularities` | array  | No       | Segment or word timestamps.                                       |

<CodeGroup>
  ```bash cURL theme={null}
  curl https://api.bianxie.ai/v1/audio/transcriptions \
    -H "Authorization: Bearer API_KEY" \
    -F "model=your-model" \
    -F "file=@speech.mp3"
  ```

  ```python Python theme={null}
  client.audio.transcriptions.create(model="your-model", file=open("speech.mp3", "rb"))
  ```

  ```javascript Node.js theme={null}
  await client.audio.transcriptions.create({ model: "your-model", file: fs.createReadStream("speech.mp3") });
  ```
</CodeGroup>

The response contains at least `text`; verbose formats may include language, duration, segments, or word timestamps. Streaming depends on model and response format.