Get API key

Audio

View as Markdown

OpenAI-compatible endpoints for text-to-speech and transcription.

PrivateMind has OpenAI-compatible endpoints that turn text into speech and speech into text.

Text to speech

POST /v1/audio/speech turns text into an audio file.

cURL
curl -s "https://api.privatemind.com/v1/audio/speech" \
  -H "Authorization: Bearer $PMIND_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "tts",
    "voice": "default",
    "input": "Hello world."
  }' \
  --output speech.mp3
  • model (required): the tts alias or a TTS model id. List with GET /v1/models.
  • voice (required): a voice id. List the voices with GET /v1/voices.
  • input (required): the text to synthesise.
  • response_format (default mp3): mp3, opus, aac, flac, wav, pcm. Engine support varies.
  • speed (default 1.0): playback speed multiplier.

The response body is the raw audio bytes.

Speech to text

POST /v1/audio/transcriptions takes an audio file and returns text.

cURL
curl -s "https://api.privatemind.com/v1/audio/transcriptions" \
  -H "Authorization: Bearer $PMIND_KEY" \
  -F "file=@call.m4a" \
  -F "model=stt"

The request is multipart form data, and curl -F sets the Content-Type for you.

  • file (required): the audio file. Supported formats: mp3, mp4, m4a, wav, webm, flac, ogg, mpga, mpeg.
  • model (required): the stt alias, or an ASR model id from GET /v1/models.
  • language: an ISO-639-1 code (en, fr, de, ...). Improves accuracy and latency.
  • prompt: priming text, such as names, jargon or context.
  • response_format (default json): json, text, srt, verbose_json, vtt.

Default json response:

JSON
{ "text": "Hello, this is a test transcription." }

verbose_json adds per-segment timestamps and the detected language. srt and vtt return subtitle files.

List voices

cURL
curl -s "https://api.privatemind.com/v1/voices?model=<tts-model-id>" \
  -H "Authorization: Bearer $PMIND_KEY"
JSON
{"object": "list", "data": [{"id": "Emily.wav", "name": "Emily"}]}

Pass an id as voice to /v1/audio/speech.

Limits

  • Transcription uploads: up to 75 MB. A larger file gets 413 before it reaches the engine. Split a long recording into segments under the cap, transcribe them in parallel and join the text.
  • TTS responses are buffered before they are sent. Keep input under a few thousand characters.

Where next

  • Models to list audio-capable models in your org.
  • Errors for the 413 upload-size response and other status codes.
  • Rate limits for budget and RPM behaviour on audio calls.