September 23, 2026
Clinical audio is transcribed more accurately: Scribe v2 Medical cuts errors by 35%
On September 23, Scribe v2 Medical reduced the word error rate on clinical audio by 35% versus the base Scribe v2. ElevenLabs fine-tuned the model to recognize drug names, anatomical terms, and pathologies.
ElevenLabs
@elevenlabs
We are releasing Scribe v2 Medical, our speech recognition model fine-tuned on clinical audio. Scribe v2 Medical reduces the word error rate (WER) on clinical audio by 35% compared with our base Scribe v2 model. It better recognizes drug names, anatomical terms, and pathology descriptions.
· 8.2K views
The base Scribe v2 recognized clinical audio without separate customization. Now ElevenLabs has released a fine-tuned version for it.
Medical vocabulary. The model should make fewer errors in drug names, anatomical terms, and pathology descriptions.
Scribe v2 Medical became a separate version in the Scribe v2 lineup for clinical audio.
