What do you need help with?
Question
Installation method
Unity Asset Store
Issue area
Hands-Free Conversation / VAD
Impact
Normal - something is not working as expected
Unity version
6000.0.59
Convai Unity SDK version
4.1.0
Target platform
Meta Quest
Where does it happen?
Both sample and custom scene
Short summary
Multiple issues related to Hands-Free Conversation, VAD, microphone muting, TTS playback, and language detection in Unity VR application.
What happened?
We are developing a VR educational application using Convai on Meta Quest and are experiencing several conversation-related issues that affect the overall user experience.
Issue 1 – Text response appears but voice never plays
Occasionally, the AI successfully generates a response, but only the text is displayed.
Observed behavior:
User speaks.
Convai processes the request successfully.
The text response appears immediately.
No voice response is played.
The conversation appears to complete successfully on the text side, but the corresponding TTS audio is missing.
Issue 2 – Long delay between text and voice
We frequently observe a noticeable delay between the appearance of the text response and the start of voice playback.
Observed behavior:
User finishes speaking.
AI text response appears quickly.
Voice playback starts several seconds later.
The delay is significant enough that users often assume the AI has stopped responding or that the response has failed.
Issue 3 – AI speech gets interrupted by minor environmental noise
The AI appears to be highly sensitive to background sounds while speaking.
Examples include:
People talking nearby
General room noise
Meta Quest microphone artifacts
These sounds often cause the AI to stop speaking or interrupt its response, even when the user is not intentionally interacting with the system.
What did you expect?
- Every AI text response should generate a corresponding voice response.
- Voice playback should begin immediately or with minimal delay after the text response is generated.
- Minor environmental sounds and background noise should not interrupt the AI while it is speaking.
04.English conversations should consistently be recognized as English without unexpected language switching or mixed-language transcriptions.
05.The conversation experience should remain stable and predictable in a typical Meta Quest classroom or training environment.
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