Artificial intelligence / news brief
ElevenLabs launches Scribe v3 batch and live speech-to-text models
- Event
- Published

ElevenLabs launched Scribe v3 for recorded audio and Scribe v3 Live for realtime transcription across 102 languages. Both are available through ElevenAPI, while the live model adds realtime speaker diarization and context-aware turn detection. Reported error reductions come from ElevenLabs tests, including an internal real-world dataset that the company says it will open later.
Why it matters
Realtime diarization and context-aware turn detection target two persistent failure points in meeting tools and voice agents, but independent tests are still needed.