TL;DR: TSR is causing weird artifacts in animated textures. Any workarounds?
More Detail:
We have a (live, virtual production) workflow where we draw video to render textures, and apply them into the scene. Unfortunately, as objects in the video move around, they do not have relevant motion vectors (because the canvas isn’t moving, or at least not in the same way as the content of the video) so the image is smudgy, smeary, and/or ghost-like. It is particularly a problem with eyes, because they can get relatively close to pixel-sized and they can move quite quickly (when people turn their heads, etc.).
Unfortunately, the alternative is no-AA or FXAA, and that negatively affects the quality of the rest of the scene.
I have some ideas for how to mitigate this. The most promising one seems to be drawing the video as a masked texture (for reflections and other depth effects) and then scheduling a custom draw, after TSR is applied, to draw the video correctly (unless the depth buffer has something occluding the video).
Any other ideas? Any advice on the double-draw idea? I know that there has been quite a bit of video playing (ex: in Fortnite).