Implemented mono depth estimation as well
I got obsessed with it and spent a lot of time coding instead of playing. But in the end I converted Depth-Anything-V2 to work for OBS.
It's getting more and more interesting to work on this project, DLSS FG works suprisingly well with estimated depth and optical flow vectors.
Maybe I'll release this after all, it's too good to keep for myself.
Comments (6)
Have you tried this injecting into like chiaki or cosplay or anything (if possible) as a workaround for the capture card
chiaki-ng uses vulkan for presentation, it's kind of a hassle
How much added latency have you perceived while doing this considering you are capturing the footage. Also will things like HDR work through this method?
What you would expect from using a capture card, barely notice any if you are using a good one.
would be keen to try it. i got a Elgato HD60 S+ so i think there would be a benifit right? It does 4K60 HDR10 Passthrough:
Well passthrough isn't what you should be looking at, it's the capture capability, cuz in the end you would need to play on OBS on windows to use framegen.