What it does:

You stream live audio over WebSocket, and we return timestamped speaker labels in real time. Providing the same accuracy as Precision-2.

• ~300 ms diarization latency

• A maximum of 8 speakers and up to 10 parallel streams is supported

• Input: 16 kHz mono audio over WebSocket, in 100ms chunks per message

• Output: start/end time, speaker label

• Built to handle noisy environments, overlapping speakers, and variable audio quality

• Free for the full beta period.

👉 Request access to the Streaming Diarization beta