Scribe v2

Scribe v2 is the speech-recognition side of ElevenLabs, built to compete directly with Whisper and the transcription tiers of the big American labs. It covers more than ninety languages and returns not just words but structure: word-level timestamps, speaker diarization telling apart who said what, and audio tagging for non-speech events. A separate real-time variant, Scribe v2 Realtime, cuts latency to roughly 150 milliseconds, which is the threshold at which live captioning and voice agents stop feeling delayed. In practice the two Scribe variants and the Eleven v3 synthesis model form the two ends of the same pipeline: hearing and speaking from one vendor.

▸Videos

No videos yet.

▸Reviews