ElevenLabs Scribe v3 Live uses conversational context to stop voice agents from cutting off callers mid-sentence
Reported by The Decoder1 source ↓
ElevenLabs released Scribe v3 for recorded audio and Scribe v3 Live for real-time transcription. Both are available through ElevenAPI and ElevenAgents; the company says Live uses conversational context to detect when a caller has finished speaking.
Key points
- Scribe v3 targets recorded audio; Scribe v3 Live is designed for real-time use
- Both models are available now through ElevenAPI and ElevenAgents
- ElevenLabs says Scribe v3 reduces word errors by about 25% versus its predecessor across 102 languages
- Artificial Analysis ranked Scribe v3 among the most accurate models it tested, but found it mid-pack on speed and price
Details
Scribe v3 Live analyzes conversational context as well as audio to judge whether a speaker has finished, rather than relying only on pause timers. The Decoder reports that the model is intended to address voice agents cutting callers off when they pause to think or read out information slowly.
ElevenLabs says Scribe v3 improves word-error rates by about 25% over its predecessor across 102 languages. The Decoder reports that Artificial Analysis found Scribe v3 among the most accurate transcription models in its benchmark, behind only an Alibaba preview model; the benchmark covered recorded audio and did not test Live’s real-time behavior or the other capabilities ElevenLabs cites. The outlet also reports that Scribe v3 was mid-pack on speed and price.
Written from coverage as of
Images
- The Decoder
Stocks in this story
Change since open: 20 Nov to 21 Nov 2026
Last refresh