Platform: Unity, Quest 3 Standalone VR Build
Issue Summary:
I’m experiencing consistent failures with ConvAI avatars approximately 10-12 minutes into the experience when deployed as a standalone build on Quest 3. The issue does not occur (or is minimal) when testing in Unity Play Mode.
Symptoms (multiple failure modes occurring):
- Avatars stop responding to user prompts entirely
- Text transcription of user speech stops working
- Avatar begins responding but audio cuts out mid-response while lip sync continues
- Feels like a memory/cache limitation being reached
Project Setup:
- Built in Unity for standalone VR (Quest 3)
- 5 total ConvAI avatars in the experience
- Each avatar exists in its own separate scene (only one ConvAI avatar per scene)
- Issue occurs only on standalone builds, not in Play Mode testing w/ PCVR
What I’ve Ruled Out:
- No custom data structures allocating to heap/cache on my end
- Issue timing is consistent (~10-12 minutes) regardless of interaction pattern
Questions:
- Could the multi-scene setup with 5 separate avatars be causing memory accumulation across scene loads?
- Are there known Quest 3 standalone limitations or memory constraints I should be aware of?
- Are there any diagnostic logs or debugging steps you’d recommend to isolate whether this is a client-side or server-side issue?
Any guidance would be greatly appreciated. Happy to provide additional technical details if needed