# Inconsistent transcripts language when using Gemini realtime model ( gemini-live-2.5-flash-native-audio )

**URL:** <https://community.livekit.io/t/inconsistent-transcripts-language-when-using-gemini-realtime-model-gemini-live-2-5-flash-native-audio/410>\
**Category:** Agents\
**Tags:** agent-development, plugin, gemini, google\
**Created:** [February 27, 2026, 9:30am UTC](https://community.livekit.io/t/inconsistent-transcripts-language-when-using-gemini-realtime-model-gemini-live-2-5-flash-native-audio/410 "2026-02-27T09:30:10Z")\
**Posts on this page:** 4\
**Page:** 1

<div class="post-metadata">

**Author:** ![Ashwin\_P](https://sea1.discourse-cdn.com/flex001/user_avatar/community.livekit.io/ashwin_p/32/518_2.png) [@Ashwin\_P](https://community.livekit.io/u/Ashwin_P)\
**Post date:** [February 27, 2026, 9:30am UTC](https://community.livekit.io/t/inconsistent-transcripts-language-when-using-gemini-realtime-model-gemini-live-2-5-flash-native-audio/410/1 "2026-02-27T09:30:10Z")

</div>

When using the Gemini realtime native audio model, the transcript occasionally returns text in non-English languages even though:

The user is speaking English

The model is configured with language=“en-US”

This happens intermittently within the same session and causes issues for downstream processing that expects a consistent English transcript

Realtime model configuration:

```auto
google.realtime.RealtimeModel(
    model="gemini-live-2.5-flash-native-audio",
    voice="Puck",
    temperature=0.3,
    language="en-US",
    location="us-central1",
    vertexai=True,
)

```

---

<div class="post-metadata">

**Author:** ![darryncampbell](https://sea1.discourse-cdn.com/flex001/user_avatar/community.livekit.io/darryncampbell/32/20_2.png) [@darryncampbell](https://community.livekit.io/u/darryncampbell)\
**Post date:** [February 27, 2026, 10:12am UTC](https://community.livekit.io/t/inconsistent-transcripts-language-when-using-gemini-realtime-model-gemini-live-2-5-flash-native-audio/410/2 "2026-02-27T10:12:11Z")

</div>

I remember this issue came up previously, but for OpenAI realtime. At the time, the solution was to provide instructions saying to “use English”, and OpenAI had documented that as a recommendation for realtime prompting ([link](https://developers.openai.com/cookbook/examples/realtime_prompting_guide) - under the ‘Language Constraint’ section). I can’t find any equivalent advice for Gemini, but it’s probably worth doing.

---

<div class="post-metadata">

**Author:** ![Ashwin\_P](https://sea1.discourse-cdn.com/flex001/user_avatar/community.livekit.io/ashwin_p/32/518_2.png) [@Ashwin\_P](https://community.livekit.io/u/Ashwin_P)\
**Post date:** [March 3, 2026, 6:14am UTC](https://community.livekit.io/t/inconsistent-transcripts-language-when-using-gemini-realtime-model-gemini-live-2-5-flash-native-audio/410/3 "2026-03-03T06:14:53Z")

</div>

Okay Thankyou..Will try this

---

<div class="post-metadata">

**Author:** ![CWilson](https://sea1.discourse-cdn.com/flex001/user_avatar/community.livekit.io/cwilson/32/135_2.png) [@CWilson](https://community.livekit.io/u/CWilson)\
**Post date:** [March 3, 2026, 8:37pm UTC](https://community.livekit.io/t/inconsistent-transcripts-language-when-using-gemini-realtime-model-gemini-live-2-5-flash-native-audio/410/4 "2026-03-03T20:37:32Z")

</div>

This is a model issue, and not related to the framework or other LiveKit functionality

[https://github.com/livekit/livekit/issues/4338](https://github.com/livekit/livekit/issues/4338)