Official provider results
Muse Voice Transcribe launch evaluation
Meta’s launch evaluation reports streaming transcription accuracy, speaker diarization, and real-time audio-perception capabilities for Muse Voice Transcribe.
Source snapshots refreshed
Five measurement lanesVOICE / S2S
Reasoning, task completion, conversation dynamics, experience, and latency stay separate.
Benchmark profile
- Scope
- Final-transcription streaming WER; average diarization error rate across AMI-IHM, AMI-SDM, and VoxConverse; multilingual and long-context capability disclosures
- Primary metric
- Word error rate and diarization error rate (lower is better)
- Owner
- Meta Superintelligence Labs
- Available evidence
- Exact figures published in the official launch post
What it measures
Conversation dynamics
Can it handle turns, interruptions, ambiguity, and state?
Voice experience
Is the exchange natural, robust, and responsive?
Latency
How long do responses, tool calls, and full tasks take?
Available results
- Final-transcription streaming WER
- 3.1%
- Average diarization error rate
- 17.5%
- Time to final transcription
- About 0.16s
#1 of 8
The other systems in Meta’s September 1 comparison range from 3.4% to 4.0%; lower is better.
#1 of 6
Average DER across AMI-IHM, AMI-SDM, and VoxConverse; the other systems range from 21.1% to 28.6%.
New Pareto point
The launch chart places Muse below the previous speed–accuracy frontier; the plotted value is approximate.
Interpretation limit
These are provider-reported launch results, not an independent end-to-end voice-agent ranking. The post does not publish enough per-dataset detail to reconstruct either aggregate from raw runs.