Agent update
Speech to text (Scribe v3)
ElevenLabs Scribe v3, launched on 10 Nov 2026, is a new speech-to-text model for recorded audio, with Scribe v3 Live for real-time audio. Both improve accuracy and speaker labels, and v3 Live can label who is speaking as the audio comes in.
How to use Scribe v3 for small business
A clinic could transcribe a session recording with clearer speaker labels, then turn the transcript into draft clinical notes for a practitioner to check. A consultant could record a client meeting, get timestamps and names on each speaker, and use that to write minutes without replaying the whole call. A front desk using ElevenAgents could switch to Scribe v3 Live so a voice agent is less likely to cut someone off when they pause mid-sentence, for example while reading out a postcode or zip code.
Scribe v3 Live is in ElevenAgents, and both models are the default in the ElevenLabs app and available in the API by changing the model ID. They run in the US by default. Isolated EU, India and Singapore environments are offered for enterprise customers with data residency needs. The announcement does not mention other regional environments beyond those. Both models are HIPAA-eligible with a Business Associate Agreement. Pricing is not listed on the launch post.
Most useful for: Executive Assistant – Take minutes and track actions after meetings, then follow up owners, Legal Secretary – Transcribe dictation, Podcast Editor – Create transcripts and captions for each episode
Frequently asked questions
Where is Scribe v3 hosted, and can I use it outside the US?
You can use the models from the ElevenLabs app, ElevenAgents and the API. They run in the US by default. ElevenLabs also offers isolated EU, India and Singapore environments for enterprise customers with data residency requirements. The announcement does not mention other regional environments.
How do I switch from Scribe v2, and is there a listed price?
The launch post does not tie Scribe v3 to a named plan or price. Scribe v3 Live is in ElevenAgents, and both models are the default in the ElevenLabs app. Existing API users switch by changing the model ID to scribe_v3 or scribe_v3_live.
Will it still cut callers off when they pause?
That is one of the problems v3 Live is built to reduce. It uses the conversation as well as silence to judge when someone has finished speaking, so a caller who pauses halfway through a number is less likely to be interrupted. It also adds speaker labels in real time, which Scribe v2 Realtime did not have.