ElevenLabs launches Scribe v3 and v3 Live speech-to-text models

發生什麼事

The company’s blog reports 27% fewer word errors than Scribe v2 across recorded and live audio (ElevenLabs’ Scribe v3 announcement). It also says the models are designed for noisy recordings and meetings with several people speaking. These are ElevenLabs’ claims; we have not tested the models (🗣 Official claims; not tested by us).