Introducing Scribe v2 Realtime

Today we're introducing Scribe v2 Realtime, the most accurate low-latency Speech to Text model, delivering live transcription in under 150 ms.

Scribe v2 Realtime: the most accurate model for live transcription

Scribe v2 Realtime sets a new standard for low-latency Speech to Text.

Designed for live use cases—voice agents, meeting assistants, and real-time captioning—it transcribes speech in under 150 ms across English, French, German, Italian, Spanish, and Portuguese, and 90 languages.

Scribe v2 Realtime delivers human-level understanding in real time, enabling natural conversation and immediate response in live environments. Scribe v2 Realtime achieves 93.5% accuracy across 30 commonly used European and Asian languages.

Key features of both models

State-of-the-art accuracy across 90 languages
Ultra-low latency with Scribe v2 Realtime (<150 ms)
Negative latency: Next word and punctuation prediction
Text conditioning: Scribe v2 Realtime continues the transcription based on the previous batch, useful when restarting a connection
Voice Activity Detection (VAD)
Manual commit: Full control over when to finalize transcript segments
Multiple audio formats: support for PCM (48kHz) and μ-law encoding
Enterprise ready with SOC 2, ISO 27001, PCI DSS L1, HIPAA, and GDPR compliance
EU and India data residency options
Zero retention mode for sensitive workloads

Build with the API

Scribe v2 Realtime is available today through the ElevenLabs API.

Explore the documentation: https://elevenlabs.io/docs/cookbooks/speech-to-text/streaming

Use Scribe v2 Realtime in ElevenLabs Agents

Deploy natural, human-sounding agents powered by Scribe v2 Realtime. Build voice assistats for support, sales, or in-product experiences that can understand and respond in real time.

Learn more: https://elevenlabs.io/agents