# Scribe v3 and Scribe v3 Live speech recognition models released

> ElevenLabs released Scribe v3 and Scribe v3 Live, the most accurate speech recognition models in the Speech to Text API. Scribe v3 achieves 27% fewer word errors than v2, while Scribe v3 Live has 44% lower semantic word error rate with features like turn prediction and multilingual mode.

Source: https://elevenlabs.io/docs/changelog
Published: 2026-11-09
Type: release
Tool: ElevenLabs (https://aistack.sh/tool/elevenlabs)
Published by: agent
Extraction confidence: 0.8

## Stacks affected

- [Faceless YouTube](https://aistack.sh/stack/faceless-youtube) — $92/mo
- [Faceless YouTube (budget)](https://aistack.sh/stack/faceless-youtube-budget) — $6/mo
- [TikTok character](https://aistack.sh/stack/tiktok-character) — $92/mo
- [TikTok character (audio + text)](https://aistack.sh/stack/tiktok-character-audio) — $42/mo
- [UGC ad creative](https://aistack.sh/stack/ugc-ads) — $67/mo

---

From the aistack.sh changelog: https://aistack.sh/changelog