real volume motion blur?
2552 3 0- Nan Guo
- Member
- 4 posts
- Joined: 8月 2014
- Offline
- jsmack
- Member
- 8043 posts
- Joined: 9月 2011
- Offline
In theory, yes.
http://jcgt.org/published/0005/01/01/paper-lowres.pdf [jcgt.org]
Built in to Houdini, no.
http://jcgt.org/published/0005/01/01/paper-lowres.pdf [jcgt.org]
Built in to Houdini, no.
- xilofoton
- Member
- 85 posts
- Joined: 7月 2007
- Offline
Just a quick theory, needs some investigation:
- use the new retime node and slow down the sim heavily ie 10x (how many substeps do you need, maybe 3-4 is enough)
- render this much longer sequence but with lower AA and other sampling settings
- still use motion blur for better results
- blend each bunch of substep frames into one to get the original length
This really needs some research, maybe you'll end up with not too much longer rendertimes and similar quality (low sampling but frame blending)… and you can also use Optix denoiser, which can work well with such soft, but noisy image content - however they not train with such motion blurred volumetric effects, so artifacts are more likely to appear…
Interesting paper btw.
- use the new retime node and slow down the sim heavily ie 10x (how many substeps do you need, maybe 3-4 is enough)
- render this much longer sequence but with lower AA and other sampling settings
- still use motion blur for better results
- blend each bunch of substep frames into one to get the original length
This really needs some research, maybe you'll end up with not too much longer rendertimes and similar quality (low sampling but frame blending)… and you can also use Optix denoiser, which can work well with such soft, but noisy image content - however they not train with such motion blurred volumetric effects, so artifacts are more likely to appear…
Interesting paper btw.
- Nan Guo
- Member
- 4 posts
- Joined: 8月 2014
- Offline
-
- Quick Links