I improved the VRChat cloth simulation, now it is actually rendered as a mesh and finally can handle self collision for proper folds. Also I moved the simulation to CRTs so it kinda works for non friends(the simulation is on the avatar!). But the collision still requires cameras.

Theoretically I could use the main camera depth for collision, so you don't need to be friended at all, but the collision quality will unfortunately be worse.

Also fun fact the simulation shader is broken on AMD GPUs hooray💀 Will need to get hands on one to actually fix this

It will never stop being funny to me that all this cool insane shit in VR is in community-driven, smaller, grassroots projects, so you'll be watching a cool-ass VR tech demo and the camera pans over to reveal a furry or anime character 99% of the time lmao

Its just how it is... But big corporations have their own idea of how things should work, which is completely misaligned with what people usually want

OMG. How do you handle self-collision? My only idea was building some expensive tree every frame or voxelizing (but insufficient grid density could make serious problems)

Since I'm limited to fragment shaders here it's quite a bit of trickery. Each particle stores the ID of the closest, but not neighboring on the cloth particle. It first is randomly initialized. The each frame I once again check a random particle in it's neighborhood [1/?]

Imagine what VRC creators would be able to do if they didn't need to use shader workarounds for everything. It's absolutely mind blowing already.

On the other hand the limitations force me to come up with rather creative solution...

That looks so good

black magic!