Docs
Audio API
All audio paths in one place: realtime conversation, voice chat, speech to text, and text to speech.
Choosing a path
Realtime runs over WebSocket, the rest over HTTP. Latency requirements differ, so pick by use case.
| Path | Protocol | Use |
|---|---|---|
| /realtime | WebSocket | Live two-way conversation |
| /chat/completions | HTTP | Voice in, voice out |
| /audio/transcriptions | HTTP | Speech to text |
| /audio/speech | HTTP | Text to speech |
Supported formats
Input accepts wav, mp3, m4a and webm. For output, choose between mp3, wav and pcm.
Limits
File uploads go up to 25MB. Split longer audio and stitch the results together.