ElevenLabs launched two speech recognition models, Scribe v3 and Scribe v3 Live, on November 10, 2026. Scribe v3 targets recorded audio, while Scribe v3 Live handles real-time audio. The company made both models available through its ElevenAPI developer platform.
Accuracy and speaker labels
Scribe v3 reduced word errors by 27% compared with Scribe v2 across 102 languages. Scribe v3 Live cut errors by 43% compared with Scribe v2 Realtime. The live model also added speaker diarization for the first time, with 39% fewer labeling errors than the next-best realtime model.
Company and enterprise support
ElevenLabs was founded in 2022 and provides voice and chat agent, creative, and developer tools. The new models are covered by SOC 2, ISO 27001, and PCI DSS Level 1 programs. Enterprise customers can use Zero Retention Mode, private deployment, and isolated EU, India, and Singapore environments.
Output controls and turn detection
Users can register up to 1,000 key terms such as brand names, acronyms, and jargon. Scribe v3 Live uses conversation context to detect when a speaker has finished, rather than silence alone. It also offers a multilingual mode and a keyterm library for large vocabularies.
Both models are available now in the ElevenLabs app and ElevenAPI. Existing users can switch by changing the model ID to scribe_v3 or scribe_v3_live. A healthcare version of Scribe v3 Live is also being released in ElevenAgents.