Company
Date Published
Author
Contact Sales
Word count
459
Language
English
Hacker News points
None

Summary

Scribe v2 Realtime, introduced by ElevenLabs, is a highly accurate, low-latency Speech to Text model designed for live applications such as voice agents and real-time captioning, offering transcription speeds under 150 ms in multiple languages including English, French, and Spanish. The model excels in challenging environments with background noise and complex information, providing features like next word prediction, automatic language detection, and manual control over transcript finalization. It supports various audio formats and complies with multiple security and privacy standards, making it suitable for enterprise use. With a 93.5% accuracy across 30 languages, Scribe v2 Realtime is available via the ElevenLabs API for building real-time, conversational voice assistants. Additionally, ElevenLabs highlighted a partnership with Sir Michael Caine to feature his voice on their app and celebrated their Impact Program, which helps veterans like Lt Col Thomas Brittingham regain their voices through technology.