# Voice/STT works in Legacy Playground but fails in the new Convai Dashboard and Unreal Engine builds

**URL:** <https://forum.convai.com/t/voice-stt-works-in-legacy-playground-but-fails-in-the-new-convai-dashboard-and-unreal-engine-builds/6577>\
**Category:** Unreal Engine Plugin\
**Tags:** language-and-speech\
**Created:** [September 20, 2026, 5:24pm UTC](https://forum.convai.com/t/voice-stt-works-in-legacy-playground-but-fails-in-the-new-convai-dashboard-and-unreal-engine-builds/6577 "2026-09-20T17:24:28Z")\
**Posts on this page:** 5\
**Page:** 1

<div class="post-metadata">

**Author:** ![theportal102](https://yyz2.discourse-cdn.com/flex004/user_avatar/forum.convai.com/theportal102/32/4976_2.png) [@theportal102](https://forum.convai.com/u/theportal102)\
**Post date:** [September 20, 2026, 5:24pm UTC](https://forum.convai.com/t/voice-stt-works-in-legacy-playground-but-fails-in-the-new-convai-dashboard-and-unreal-engine-builds/6577/1 "2026-09-20T17:24:28Z")

</div>

### What do you need help with?

Bug report

### Installation method

Not sure

### Issue area

Push-to-Talk

### Impact

Blocker - I cannot continue

### Unreal Engine version

UE5.6.1

### Convai Unreal Engine Plugin version

4.0.0-beta.29.1

### Character ID

a2f4410e-feb9-11f0-99ca-42010a7be027

### Project setup

Blueprint

### Target platform

Unreal Editor

### Where does it happen?

Custom project

### Short summary

hold t to talk not detect voice

### What happened?

Voice/STT suddenly stopped working in the new Convai Dashboard and in Unreal Engine integrations.

The same character works correctly with voice in the Legacy Playground.

However, voice input does not work in:

the new Convai Dashboard;  
several Unreal Engine projects that previously worked correctly;  
previously packaged Windows builds that also worked correctly before.

Text input works normally.

In the new Dashboard, Voice Mode shows “Connected” and “Real-time audio streaming”, but spoken input is not processed into a response.

In Unreal Engine, Push-to-Talk also no longer produces a response.

I tested this on more than one computer, with the same result.

The microphone itself is working correctly because voice conversation works normally with the same character in the Legacy Playground.

Account quota is not exhausted: 366 / 3000 interactions.

This appears to be related to the current realtime/STT pipeline used by the new Dashboard and Unreal integration.

Could you please check whether final-user-transcription or the corresponding STT result is being generated correctly for these sessions?

I can provide Unreal logs, screenshots, plugin version, and additional diagnostic information if needed.

### I checked the relevant documentation, resources, and similar forum posts before posting.

on

### I confirm that I did not include API keys, tokens, passwords, secrets, private character data, private conversation logs, or customer data.

on

---

<div class="post-metadata">

**Author:** ![theportal102](https://yyz2.discourse-cdn.com/flex004/user_avatar/forum.convai.com/theportal102/32/4976_2.png) [@theportal102](https://forum.convai.com/u/theportal102)\
**Post date:** [September 20, 2026, 5:29pm UTC](https://forum.convai.com/t/voice-stt-works-in-legacy-playground-but-fails-in-the-new-convai-dashboard-and-unreal-engine-builds/6577/2 "2026-09-20T17:29:22Z")

</div>

**Update: Unreal log confirms an STT turn-finalization failure.**

Push-to-Talk itself is working correctly:

`ConvaiPlayerLog: Started talking`

On release, the client sends:

`ConvaiSubsystemLog: Sent force-user-stopped-speaking`

The server successfully mutes STT:

`event_type: stt-toggle`  
`message: STT muted`  
`status: success`

But immediately afterward the server returns:

`event_type: force-user-stopped-speaking`  
`reason: stt_finalization_failed`  
`status: failed`  
`stop_id: null`  
`message: User speech finalization failed`

Unreal then logs:

`Server response 'force-user-stopped-speaking' failed: User speech finalization failed`

No final user transcription is produced and therefore the character never receives the user turn.

Voice works correctly for the same character in Legacy Playground, while it fails in the New Dashboard and Unreal projects/builds.

This appears to be a realtime STT turn-finalization issue rather than a microphone or Unreal input issue.

Could you please check the backend/native realtime client handling of `force-user-stopped-speaking` and `stt_finalization_failed`?

---

<div class="post-metadata">

**Author:** ![K3](https://yyz2.discourse-cdn.com/flex004/user_avatar/forum.convai.com/k3/32/2611_2.png) [@K3](https://forum.convai.com/u/K3)\
**Post date:** [September 21, 2026, 8:09am UTC](https://forum.convai.com/t/voice-stt-works-in-legacy-playground-but-fails-in-the-new-convai-dashboard-and-unreal-engine-builds/6577/3 "2026-09-21T08:09:23Z")

</div>

Could you please try again?

---

<div class="post-metadata">

**Author:** ![theportal102](https://yyz2.discourse-cdn.com/flex004/user_avatar/forum.convai.com/theportal102/32/4976_2.png) [@theportal102](https://forum.convai.com/u/theportal102)\
**Post date:** [September 21, 2026, 9:02am UTC](https://forum.convai.com/t/voice-stt-works-in-legacy-playground-but-fails-in-the-new-convai-dashboard-and-unreal-engine-builds/6577/4 "2026-09-21T09:02:34Z")

</div>

Thank you! Is working great.

---

<div class="post-metadata">

**Author:** ![K3](https://yyz2.discourse-cdn.com/flex004/user_avatar/forum.convai.com/k3/32/2611_2.png) [@K3](https://forum.convai.com/u/K3)\
**Post date:** [September 21, 2026, 9:20am UTC](https://forum.convai.com/t/voice-stt-works-in-legacy-playground-but-fails-in-the-new-convai-dashboard-and-unreal-engine-builds/6577/5 "2026-09-21T09:20:56Z")

</div>


