I noticed my scene captures/render targets still use the raster methods so they look different from everything else. Is there a reason for this or is this just an oversight?
I wouldn’t call it an oversight, raytracing from scenecaptures would be REALLY expensive.
I don’t think he means recapturing during runtime, but when you refresh them in editor?
even at runtime, one might want to capture one frame.
or maybe someone has a very lightweight game and can afford always-on scenecaptures with raytracing. the performance cost argument of scenecaptures is no different whether it uses raytracing or not :rolleyes:
Yes, this is more for recording images for videos than having them used realtime in a game.
It’s not an oversight, raster is just plain faster for scene captures. And you can already re-do scene capture like stuff in realtime, several games shipped this gen doing that. Spiderman among them, instead of saving their cubemaps they actually just render them as you get close enough. There was even an entire level in Call of Duty (Infinite Warfare?) that used a bunch of realtime captures for dynamic GI.
To be clear what I wanted to know is if DXR for scene captures (2d and cube) is a planned feature or not. Speed is not important, if needed I would even pause the game for a few frames if denoising needed it before saving the render targets to disk.
Since DXR itself is a new feature DXR for scene captures may not be the highest priority but it would still be interesting to see if it’s planned or not.
Probably no, because, again, why would it be? There’s zero need for it, it wouldn’t do anything scene capture doesn’t do today, and it would just do it slower. There no benefit whatsoever to this.
For games there’s a smaller benefit and it’s probably not viable until even faster cards are available but for linear media there’s a huge benefit, for example VR movie rendering for archviz. ODS or just normal scene capture cubes are way faster than the panoramic plugin and would look really good with DXR.