Recently, I discovered Deepgram’s Multichannel transcription feature:
According to the documentation, it transcribes each audio channel independently.
Has anyone implemented this with LiveKit? How can I integrate it into my LiveKit voice agent? Is this supported, and if so, what’s the recommended approach?
I see a channels listed under the model parameters for Nova models here:
However, its designed for scenarios where different speakers are on different channels, which is uncommon in realtime voice AI.
Are you looking to identify and label separate speakers in an audio stream? In which case, you want to use speaker diarization (lower down on the same page I linked above)