Voice-agent and live-caption builders who need fast transcription; at $0.39 an hour it costs more than the batch model.
Kenning's take
What it does
Scribe v2 Realtime transcribes live speech in about 150 milliseconds across 90+ languages, for voice agents, meeting assistants and live captions. It predicts the next word, detects language switches mid-conversation and supports telephone audio formats. ElevenLabs released it on 11 November 2025 through its API and in ElevenLabs Agents.
- About 150ms latency
- 90+ languages with automatic language detection
- Voice activity detection and manual commit
- PCM and mu-law audio support
- Built into ElevenLabs Agents
Who it suits. Developers building voice agents, call tools and live captioning.
Key facts
- Category
- AI models
- Made by
- ElevenLabs
- Based in
- United States
- Released
- 11 November 2025
- How to get it
- through the ElevenLabs API and inside ElevenLabs Agents
- Works on
- API
- In the UK
- Yes
- Free plan
- 2 hours 30 minutes on the Free plan
- API price
- $0.39 per hour of audio; keyterm prompting $0.08 per hour, transcript editing $0.12 per hour
FAQ
When was Scribe v2 Realtime released?
ElevenLabs released it on 11 November 2025.
How fast is Scribe v2 Realtime?
ElevenLabs puts latency at about 150 milliseconds.
How much does Scribe v2 Realtime cost?
$0.39 per hour of audio through the API.
Facts from the maker's own pages and the sources listed. Checked 2026-10-11.
Similar tools
Eleven v3
Expressive ElevenLabs text-to-speech model with audio tags and multi-speaker dialogue
Eleven v4
ElevenLabs' most emotive text-to-speech model, with a fast Turbo version for voice agents
Scribe v2
ElevenLabs' speech-to-text model for accurate batch transcription, subtitles and captions in 90+ languages
ACE-Step 1.5
Open-source music generation model that writes full songs with lyrics and runs on ordinary consumer GPUs.