Skip to main content

Choose which vendor hears

PUT 
/api/voice/transcription

The speech-to-text half. Live call captions, voice notes in chat and the speaker-labelled two-leg call dialog all resolve through this one selection.

A SEPARATE verb from the announcement voice, because the two are different markets — the vendor with the best Turkish voice is not automatically the one with the best Turkish ears — and a tenant changing one must not silently re-post the other.

The vendor list is narrower than the voice picker's: only openai and deepgram are wired for pre-recorded transcription. Offering a vendor that merely has a speech adapter would be a choice that throws on the tenant's first voice note.

Sending provider: null returns to the platform's default.

Request​

Responses​

The saved selection.