Request

The request body is multipart/form-data with two parts, audio and config.

audio partCopy Anchor Link

The audio can be either a WAV file or raw PCM.

• WAV (audio/wav): sample rate and channel count are read from the header.
• Raw PCM (audio/pcm): int16 little-endian, no header. Requires sampleRate and channels in the config.

The declared content type alone decides the format; the bytes are never inspected. Any other content type, including application/octet-stream, is refused with 400.

Stereo is down-mixed to mono, and any sample rate is resampled for you. No client-side conversion is needed.

config partCopy Anchor Link

A JSON object with the following fields:

• language (required): the spoken language. Supported: fr, en, pt, de, es, it.
• sampleRate (required for raw PCM): source rate in Hz. Ignored for WAV.
• channels (required for raw PCM): channel count. Ignored for WAV.
• context (optional): contextual biasing to apply when transcribing the audio. A request carries one audio file with no session around it, so this one field does both jobs the Real-time API splits across sessionContext and turnContext: the facts that ground the recording, and the values you expect to hear in it. It contains:
  • facts (optional): objects with name, value, and an optional description.
  • hints (optional): objects with type set to values, a values list, and an optional name and description.

Example config:

{
  "language": "fr",
  "context": {
    "facts": [
      { "name": "customerName", "value": "Jean Dupont" }
    ],
    "hints": [
      {
        "type": "values",
        "name": "carModel",
        "values": ["208", "2008", "3008", "5008"]
      }
    ]
  }
}

Example config (raw PCM):

{
  "language": "fr",
  "sampleRate": 8000,
  "channels": 2
}

See the Transcribe a file quickstart for complete curl and Python clients.