AI BriefingElevenLabs launches Scribe v3 and v3 Live with 27% and 43% error reductions
·2026.11.10 21:00
Key point
Scribe v3 Live reduces realtime semantic errors by 43% and adds speaker diarization, while Scribe v3 improves batch accuracy in every tested language.
Details
- ElevenLabs launches Scribe v3 for recorded audio and Scribe v3 Live for realtime, reducing word errors by 27% and semantic word errors by 43% respectively compared to Scribe v2 versions.
- Scribe v3 achieves the lowest diarization error rate among tested models, with 17% fewer errors on recordings with three or more speakers.
- Scribe v3 Live introduces speaker diarization to realtime transcription, achieving a 3.1% error rate in English, significantly lower than competitors' 5.9% to 7.3%.
- The models support up to 1,000 key terms for prompting and include a keyterm library for larger vocabularies, with improved turn detection that uses conversation context to avoid cutting off speakers.
- Enterprise features include isolated environments in the EU, India, and Singapore, Zero Retention Mode, and compliance with SOC 2, ISO 27001, and PCI DSS Level 1.
- A healthcare-optimized version of Scribe v3 Live is available in ElevenAgents for medical conversations.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
ElevenLabs launches Eleven Realtime v1 Mini and Max speech-to-speech modelsElevenLabs Introduces ElevenAI Desktop App for Dictation and Meeting Notes