ElevenLabs launches Scribe v3 and Scribe v3 Live for speech recognition

Scribe v3 handles recordings and Scribe v3 Live handles streaming audio. In company evaluations, v3 made 27% fewer word errors than v2 across 102 languages, while Live made 43% fewer errors than v2 Realtime. The live model adds speaker diarization and improves turn detection.

Visit original source