ElevenLabs updates Scribe v2 with entity redaction and cleaner transcripts
ElevenLabs updated Scribe v2 with four new features including automatic sensitive data redaction, cleaner transcripts, and expanded keyterm support.
It’s been a busy day for AI speech. Microsoft launched MAI-Transcribe-1 claiming the top spot on the FLEURS benchmark, and Fish Audio added an STT tool built around emotion-aware transcription. ElevenLabs isn’t sitting still either.
The company pushed four updates to Scribe v2 today, its speech-to-text model. The biggest addition is entity redaction, which automatically strips sensitive information from transcripts before anything gets stored.
That means names, credit card numbers, Social Security numbers, and other personal data are removed during transcription itself, not after. Teams in healthcare, finance, and customer support are the obvious target here.
The new features:
- Entity redaction: Three output options — full redaction with [REDACTED], category labels like [CREDIT_CARD], or enumerated labels like [CREDIT_CARD_1]
- Indic-English code-switching: Scribe now keeps English words in Latin script when speakers mix Hindi, Telugu, or Kannada with English, instead of transliterating them into Indic scripts
- No Verbatim mode: Automatically removes filler words, repeated phrases, and stuttering for cleaner output without manual editing
- Keyterm prompting: Expanded from 100 to 1,000 terms per transcript, with context-aware matching rather than blind insertion
One pricing note on the keyterms update. Requests with more than 100 keyterms carry a minimum billable unit of 20 seconds.
All four features are available now in the Scribe v2 API and UI.
Source: ElevenLabs Blog
Get the core business tech news delivered straight to your inbox. We track AI, automation, SaaS, and cybersecurity so you don't have to.
Just read what you want, and be done with it.
