Speaker Diarization And Channels
Deepgram Speech AI Platform · Deepgram
Speech AI Platform — Deepgram (Nova STT, Aura TTS, Voice Agent)
Deepgram evals — Speaker Diarization & Channels (relift v3 InfraRed)
About Deepgram
Deepgram is a speech-AI platform offering streaming and batch speech-to-text (Nova), Aura text-to-speech, speaker diarization, redaction, and smart formatting across 30+ languages — used by voice-agent platforms, contact centers, and media teams.
Sample tests· showing 3 of 9
| # | Input | Expected behavior | Check |
|---|---|---|---|
| 01 | Operator transcribes a mono recording of a 3-speaker meeting with diarize=true. | Read per-word speaker labels at results.channels[0].alternatives[0].words[i].speaker. Speaker labels are integers (0, 1, 2, ...) assigned per request and not portable across recordings. Group by speaker for utterances. | Pass / FailAi Platformhigh |
| 02 | Operator wants utterance-level chunks each carrying a single speaker label. | Combine utterances=true with diarize=true; utterances inherit the speaker label of the contained words. When words within one utterance carry different speaker labels, split client-side at the label boundary or accept the majority-vote label depending on UX. | Pass / FailAi Platformmedium |
| 03 | Operator wants to pin the diarization version to insulate against rollouts. | Use diarize_model=<version> (e.g., 'latest' or a pinned variant) to control which model performs diarization. Pin in production for regression stability. Do NOT set both diarize=true and diarize_model on the same request — the API rejects the combination per docs. | Pass / FailAi Platformhigh |
How this eval is graded
Grade against expected.ideal_behavior and expected.rubric. Per-criterion pass requires mean >= 4.0 and no criterion below 3.
Rubric criteria
- Deepgram
- Ai Platform
- Speaker Diarization And Channels
Recommended for
Works with
Related evals
Run this eval in your workspace
Connect your data, configure thresholds, and review results with your team.