Real-time speaker diarization
Identify and separate speakers as the conversation unfolds. Enable per-session with a single flag, no extra configuration needed.
Drop-in integration
A single API call is all it takes. Plug into your existing audio stack and go from sign-up to production in minutes, not days.
Value without the wait: Provide insight, assistance and understanding, to the user immediately, without waiting until the end of the audio.
Industry-leading accuracy
Best-in-class speaker diarization, even with overlapping speech or rapidly switching turns. The speed-accuracy tradeoff is no longer a constraint.
Sub-300 ms latency
Speaker labels arrive fast enough for live captioning and real-time agent assist. Tested against noisy, real-world audio conditions, not just clean studio recordings.
Use cases
What you can do with it
~300 ms
diarization latency
16 kHz
mono audio over WebSocket, in 100 ms chunks per message
Output
start time, end time, speaker label
Up to 8
speakers and 10 parallel streams
Up to 5 hours
duration per stream
WebSocket API
for easy integration with your existing STT pipeline