---
title: "LiveKit Turn Detector v1.1: end-of-turn detection that reads and hears"
date: "2026-11-20"
authors: "Leigh Weston, Chenghao Mou"
tags: "product"
---

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

LiveKit Turn Detector v1.1 detects the end of a user's turn 15–34% sooner than v1, at the same false-interruption rate. It ships with v1.1-mini, an open-weight version that runs on CPU.

Like [v1](/blog/solving-end-of-turn-detection), v1.1 listens to the audio of the user's turn. It also reads the agent's last message as text. v1.1-mini decides from audio alone. Both models improve on two cases that are hard for end-of-turn detection: short replies and non-native English speakers.

## Short turns

When an agent asks a direct question, users often answer in a few words, such as a number, a name, or a short "uh-huh". But the same words can start a much longer answer to other questions:

- Agent: "How many guests?"

- User: "Seven."

- Agent: "What is your phone number?"

- User: "Seven… \[pause]"

To tell the two apart, v1.1 also reads the agent's last message.

## Non-native English speakers

Non-native speakers often pause in the middle of a sentence to find a word. A model trained mostly on native speech can take that pause for the end of the turn, and the agent cuts in too early. Both models are trained on additional non-native English speech from many accents, so they wait through these pauses.

## Benchmark results

To measure progress on these two cases, we're adding two English evaluation sets to [eot-bench](https://github.com/livekit/eot-bench): 1,088 short user turns and 800 turns from non-native English speakers. They join the original 400-turn English set and up to 400 turns in each of the 13 other languages. Use them to test any end-of-turn model on these cases, with the same harness behind the results below.

Latency is how long the model waits after the user stops speaking before it ends the turn. The cut-off budget caps false interruptions, where the agent cuts in before the user is done, at 5% or 10%. Lower latency is better.

### English

Cut-off budget 5%

<ComparisonTable highlight={['v1.1', 'v1.1-mini']} wide>
  | Eval set                                                              | v1.1                                                    | v1       | v1.1-mini                                               | v1-mini  |
  | --------------------------------------------------------------------- | ------------------------------------------------------- | -------- | ------------------------------------------------------- | -------- |
  | Original set <div className="text-fg3 text-sm">400 turns</div>        | **462 ms** <div className="text-fgSuccess">−15%</div>   | 543 ms   | **812 ms** <div className="text-fgSuccess">−24%</div>   | 1,070 ms |
  | Short turns <div className="text-fg3 text-sm">1,088 turns</div>       | **1,096 ms** <div className="text-fgSuccess">−23%</div> | 1,427 ms | **1,237 ms** <div className="text-fgSuccess">−24%</div> | 1,634 ms |
  | Non-native speakers <div className="text-fg3 text-sm">800 turns</div> | **700 ms** <div className="text-fgSuccess">−19%</div>   | 867 ms   | **999 ms** <div className="text-fgSuccess">−19%</div>   | 1,228 ms |
  | All <div className="text-fg3 text-sm">2,288 turns</div>               | **810 ms** <div className="text-fgSuccess">−19%</div>   | 996 ms   | **1,124 ms** <div className="text-fgSuccess">−17%</div> | 1,352 ms |

  Percentages compare each model with its predecessor.
</ComparisonTable>

Cut-off budget 10%

<ComparisonTable highlight={['v1.1', 'v1.1-mini']} wide>
  | Eval set                                                              | v1.1                                                  | v1     | v1.1-mini                                             | v1-mini  |
  | --------------------------------------------------------------------- | ----------------------------------------------------- | ------ | ----------------------------------------------------- | -------- |
  | Original set <div className="text-fg3 text-sm">400 turns</div>        | **250 ms** <div className="text-fgSuccess">−15%</div> | 295 ms | **504 ms** <div className="text-fgSuccess">−28%</div> | 698 ms   |
  | Short turns <div className="text-fg3 text-sm">1,088 turns</div>       | **635 ms** <div className="text-fgSuccess">−32%</div> | 932 ms | **893 ms** <div className="text-fgSuccess">−19%</div> | 1,100 ms |
  | Non-native speakers <div className="text-fg3 text-sm">800 turns</div> | **360 ms** <div className="text-fgSuccess">−31%</div> | 523 ms | **676 ms** <div className="text-fgSuccess">−24%</div> | 884 ms   |
  | All <div className="text-fg3 text-sm">2,288 turns</div>               | **414 ms** <div className="text-fgSuccess">−34%</div> | 632 ms | **738 ms** <div className="text-fgSuccess">−20%</div> | 922 ms   |

  Percentages compare each model with its predecessor.
</ComparisonTable>

### Multilingual

<ComparisonTable highlight={['v1.1', 'v1.1-mini']} wide>
  | Cut-off budget | v1.1                                                  | v1     | v1.1-mini                                             | v1-mini |
  | -------------- | ----------------------------------------------------- | ------ | ----------------------------------------------------- | ------- |
  | ≤5%            | **521 ms** <div className="text-fgSuccess">−17%</div> | 631 ms | **715 ms** <div className="text-fgSuccess">−18%</div> | 867 ms  |
  | ≤10%           | **329 ms** <div className="text-fgSuccess">−19%</div> | 406 ms | **492 ms** <div className="text-fgSuccess">−22%</div> | 629 ms  |

  The mean of 13 languages, with up to 400 turns each. Percentages compare each model with its predecessor.
</ComparisonTable>

At the same cut-off budget, v1.1 cuts latency by 15–34% on English and by 17–19% across the 13 other languages, compared with v1. v1.1-mini cuts latency by 17–28% on every set, compared with v1-mini.

The [eot-bench leaderboard](https://livekit.com/benchmarks/eot-bench) ranks models on each set. v1.1 ranks first on every English set at both 5% and 10% false-interruption rates. Most end-of-turn models don't cover all 13 other languages, so the leaderboard also has an interactive heatmap of per-language results. v1.1 has the lowest latency across all languages at both rates.

## Availability

LiveKit Turn Detector v1.1 is free for [agents hosted on LiveKit Cloud](/products/agent-cloud-deployment), and self-hosted agents can use it through [LiveKit Inference](/products/inference). If you already use the LiveKit turn detector through LiveKit Inference, you get v1.1 with no code change, and it decides from audio alone. Upgrade to Agents 1.8.12 (Python) or 1.10.0 (Node.js) so your agent also sends its last message and the user's turn length, which v1.1 uses for short replies.

<TabbedCodeBlock>
  <TabbedCodeBlock.File label="Python">
    ```python
    from livekit.agents import AgentSession, TurnHandlingOptions, inference

    session = AgentSession(
        turn_handling=TurnHandlingOptions(
            turn_detection=inference.TurnDetector(),
        ),
        # stt=..., llm=..., tts=...
    )
    ```
  </TabbedCodeBlock.File>

  <TabbedCodeBlock.File label="Node.js">
    ```typescript
    import { inference, voice } from '@livekit/agents';

    const session = new voice.AgentSession({
      turnHandling: {
        turnDetection: new inference.TurnDetector(),
      },
      // stt, llm, tts, ...
    });
    ```
  </TabbedCodeBlock.File>
</TabbedCodeBlock>

v1.1-mini runs on CPU, decides from audio alone, and covers the same 14 languages as v1.1. It ships with open weights in the [Agents SDKs](/products/agents-sdks) from the same versions, under the same [LiveKit Model License](https://github.com/livekit/agents/blob/main/MODEL_LICENSE) as v1-mini.

## Get started

You can use v1.1 now:

- [Turn detection in LiveKit Agents (docs)](https://docs.livekit.io/agents/logic/turns/turn-detector/#audio-turn-detector)
- [eot-bench on GitHub](https://github.com/livekit/eot-bench)
- [eot-bench datasets on Hugging Face](https://huggingface.co/datasets/livekit/eot-bench-data)
- [Voice AI quickstart](https://docs.livekit.io/agents/start/voice-ai/)

Questions, benchmarks, or pull requests? Join us in the [LiveKit community](https://community.livekit.io/).