The problem is in the way the high resolution screenshot function is implemented: it splits up the screen in a number of pieces and renders each piece separately. This means that there is no complete screen buffer that SSR can read from. Effectively this would mean that you would only get very close objects reflected and ugly edges at the seams of the stitched images.
TAA is also incompatible with the tiled approach as it depends on data from previous frames. By jumping the camera around to create the different tiles there is no data for TAA to work with. Although I suppose they could fix this by rendering each tile a number of times, although that would slow down the function a lot.
Anyway, on to possible solutions:
- If you have an Nvidia card you can use Dynamic Super Resolution (DSR) to effectively get a 4x resolution increase. The image gets downscaled for display on your screen but is full resolution internally, so screenshots will be at a resolution beyond what your monitor can display.
- Set screenpercentage beyond 100 in your post process volume. This also super samples the image, although it won’t affect the output resolution, it will improve the image quality and you can combine it with regular high res screenshots or the previously mentioned DSR.
Those methods will have a big performance impact, but if it’s just for screenshots that shouldn’t be an issue.