This makes me suspect a performance issue in the Babylon hit-test implementation. If I place an object without enabling hit-test, it stays perfectly stable.
One more detail: even if I disable hit-test after placing the object, the session remains sluggish, as if something under the hood is still doing work.
Any insights or suggestions would be greatly appreciated. Thanks!
I’m not able to reproduce this on my Quest 3, yet. The placed objects are stable for me. They do not stutter or drift, and I’m not seeing any performance issues with hit-test enabled.
Maybe you have a lot more WebXR planes being detected than me? I added plane id and count logging in https://playground.babylonjs.com/#KDWCZY#1044 . Try it out and see how many planes are detected. I’m getting 8 on my headset.
FYI I haven’t forgotten about this. I did some testing last week and it looks like the model being inserted is fairly complicated and is causing the FPS to drop, which may explain the stuttering you’re seeing. I plan on looking into this further soon to see why the FPS drops so much and to see if there’s anything we can do to speed it up on our end.
Great to hear you’re still investigating this!
I realize the model itself is fairly heavy, but since the WebXR samples were running smoothly, I was hoping there might be some small tweaks or optimizations on the Babylon side to make things a bit smoother.
Let me know if there’s anything I can do on my side to help with the investigation.
Ya, I looked at the WebXR example code briefly and I don’t think the PBR shaders are doing much. Babylon’s PBR shaders are a lot better, and as a result they are more expensive on the GPU.