Content Deep Dive

LiveKit Turn Detector v1.1: end-of-turn detection that reads and hears

Blog post from LiveKit

Post Details
Company
Date Published
Author
Leigh Weston
Word Count
850
Company Posts That Month
8
Language
English
Hacker News Points
-
Post removed?
No
Summary

LiveKit Turn Detector v1.1 identifies the end of a user’s turn 15–34% faster than v1 at the same false-interruption rate, while the open-weight v1.1-mini runs on CPU and makes decisions from audio alone. The full v1.1 model also considers the agent’s last message, helping distinguish brief answers from pauses before longer replies; both models are trained on additional non-native English speech to reduce premature interruptions when speakers pause mid-sentence. New eot-bench evaluation sets cover 1,088 short English turns and 800 turns from non-native English speakers, alongside existing English and multilingual tests. Results show v1.1 leading across all English evaluation sets and languages tested, while v1.1-mini also improves on its predecessor across every set. v1.1 is available for LiveKit Cloud agents and through LiveKit Inference, with newer Agents SDK versions enabling the contextual information it uses; v1.1-mini is available with open weights in the SDKs and supports the same 14 languages.

Trends Found in this Post
Trend Post Mentions Total Month Mentions Posts Companies MoM
Voice AI 2 1,653 202 45 -40%
LLM 1 3,674 823 199 -44%
Use This Data

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.