Speech to Text

Cube AI supports OpenAI-compatible speech-to-text transcription for audio inputs, depending on the enabled backend and available models.

⚠️ Not all Cube AI deployments enable speech-to-text by default.

Endpoint

POST /proxy/{domain_id}/v1/audio/transcriptions

Example Request

curl -k https://localhost/proxy/<domain_id>/v1/audio/transcriptions \
  -H "Authorization: Bearer <pat>" \
  -H "Content-Type: multipart/form-data" \
  -F "[email protected]" \
  -F "model=whisper"

Request Parameters

• file — Audio file to transcribe (e.g. WAV, MP3)
• model — Speech-to-text model (for example whisper)

Response

Returns a transcription result compatible with the OpenAI API format.

Notes

• Audio is processed inside a Trusted Execution Environment (TEE)
• Availability depends on backend configuration
• Large audio files may increase latency