What it does
Deepgram is the speech-to-text provider Custral uses to transcribe voice call recordings. There is no per-workspace setup. Transcription is built in. The integration appears as a configuration toggle on each voice channel.Connect
Transcription is enabled and managed by Custral for you. There’s nothing to connect. Turn it on or off per voice channel (see below).Configure
Per voice channel:- Transcription on / off: turns Deepgram on for that channel.
- Speaker diarization: separate the two sides of the call. On by default.
- Language: set per channel; defaults to English.
Resources synced
Limitations
- Transcription happens after the call ends; it is not real-time.
- Long calls (>1 hour) are split into segments; the transcript reassembles them but speaker labels can drift across segment boundaries.