Model releaseImportance 3 of 5
ElevenLabs releases Scribe v3 and Scribe v3 Live speech-to-text models
ElevenLabs announced Scribe v3 for recorded audio and Scribe v3 Live for realtime audio. The company states both are its most accurate speech-to-text models, with 27% fewer word errors than Scribe v2.
Why it matters
Builders using speech-to-text for transcription, captions, or voice interfaces may see fewer transcription errors by switching to the new models, which could reduce manual correction and downstream errors in workflows.
Who should care
Ecommerce brands using voice or video content, podcast and video creators, customer support teams, developers building voice agents or transcription features.
What you can do
Run a sample of your own recorded and live audio through Scribe v3 and Scribe v2 and compare word error rates before changing production transcription.