Closed mrjustaguy closed 5 months ago
The blue motion vectors are derived from the camera and the depth value. Is it possible it's related to reverse depth? I'd try before and after this commit https://github.com/godotengine/godot/commit/69a4ff899618cc40bcf5aed2130d4cb2e3c021d6.
~I can't reproduce this issue. I've tested a basic scene with a plane mesh with Dev 5, Dev 6 and current master.
Can you upload an MRP that reproduces the issue? There must be something about your mesh that is impacting this
Notably, my motion vectors are also always yellow rather than blue. So I'm clearly not even triggering the same codepath you are~
Edit: I figured it out. This is only reproducible when using FSR2. TAA is not affected
Yeah my bad, TAA is just it's regular old broken self due to it's own issues and hasn't regressed.. updated OP
The blue motion vectors are derived from the camera and the depth value. Is it possible it's related to reverse depth? I'd try before and after this commit 69a4ff8.
I've tested the commit before and after reverse Z and I can confirm this issue was introduced during that, and it seems to affect the derivation of the motion vectors from depth in particular that is used for FSR2. I'm not entirely sure why it breaks and how to fix it yet, but that's the starting point for this.
That is independent of the math. I opened this PR that seemingly fixes the issue and it makes enough sense to me at least: https://github.com/godotengine/godot/pull/91799
Tested versions
Happens 4.3 Dev 6 Doesn't in 4.3 Dev 5 and earlier
System information
Windows 11, i3-10105f, RX6600 24.4.1, Forward+
Issue description
Derived Motion Vectors are Broken resulting in broken FSR2 Behaviors When zooming out these are the results
Expected
Steps to reproduce
Enable FSR and move around, Looking at the Motion Vectors
Minimal reproduction project (MRP)
N/A