September 22, 2026
Medical records will have fewer errors: Scribe v2 Medical cuts WER by 18%
On September 22, ElevenLabs released Scribe v2 Medical: on clinical audio recordings, the model reduces WER by 18% compared with the base Scribe v2. It recognizes drug names, anatomical terms, and pathologies more accurately.
ElevenLabs
@elevenlabs
We are releasing Scribe v2 Medical, our speech-to-text model fine-tuned on clinical audio recordings. Scribe v2 Medical reduces word error rate (WER) on clinical audio recordings by 18% compared with the base Scribe v2 and recognizes drug names, anatomical terms, and pathologies more accurately.
· 3.6K views
Previously, the base Scribe v2 handled clinical recordings without dedicated adaptation. Now ElevenLabs has added a model fine-tuned on this type of audio.
Scribe v2 Medical has become a separate Scribe v2 variant for clinical audio recordings.
