Scribe v2
Scribe v2 improves recognition of long recordings, pauses and changes in delivery. It detects language within a recording and returns structured transcript data with speaker labels, word timing and non-speech events.
Transcript controls
The April update added automatic redaction and a No Verbatim mode that removes fillers, repeated phrases and stuttering. It also keeps English words in Latin script when speakers mix English with Indic languages. Redaction is automated and may miss sensitive information.ElevenLabs: April 2026 Scribe v2 update
Practical guidance. Choose verbatim output when the exact spoken record matters. Choose cleaned text for readable notes only after accepting that some speech details will be removed.ElevenLabs: April 2026 Scribe v2 update
Specifications & access
| Detail | Value and conditions |
|---|---|
| Launch date | |
| Access | API, Provider product |
| Model identifiers | scribe_v2 |
| Keyterm guidance | Up to 1,000 words or phrases after the April 2026 updateElevenLabs: April 2026 Scribe v2 update |
Related models
- Scribe v2 Medical
ElevenLabs · Scribe
Clinical-audio fine-tune of Scribe v2 for terminology and dictated notes.
Launch date API - Scribe v2 Realtime
ElevenLabs · Scribe
Streaming recognition model with partial and committed transcripts.
Launch date API · Provider product - Scribe v1
ElevenLabs · Scribe
Deprecated first Scribe recognition model with speaker labels and audio-event tags.
Launch date Deprecated API; check access detailsAPI