I mean surely you can understand that implicit changes like this are completely undiscernable by the user and are far worse than having to set a dropdown once in a while, right ? that seems obvious
Hard disagree.
Found this:
I am curious what Blender devs are planning towards improved caustics in Cycles and Eevee.
Btw, Excellent work by the addon dev.
Path guiding is already in master (CPU only at the moment) and that can help with caustics, although it wasnāt designed specifically for that purpose.
There is a specific thread where people have been testing it and posting results:
As I remember, Maya have -on Preferences- CM Rules for texture inputs
May be will be nice to have something similarā¦
At last! Fantastic news! Yaayā¦!
Hopefully it wonāt take too long for Nvidiaās and AMDās GPU support and weāll be there!
A new patch fixing the primary reason why the new hair shader seems inferior to the old one in numerous cases. It turns out the importance sampling was broken.
#115241 - WIP: Fix #114780: Cycles: Importance sampling correction - blender - Blender Projects
Do a blink comparison of the images and you will find just how much better the shader will be once this is committed.
Awesome. Thanks for posting.
Since I doubt most people here have an Intel GPU to test, my preliminary results with this are that it is blazing fast and causes Blender to have some artifacts across the window when denoising happens https://projects.blender.org/blender/blender/issues/115197 which should get fixed pretty soon, hopefully. It is an alpha after all.
If we ignore this bug, I can now have denoising always on without any super noticeable performance reduction, compared to big lag when having it on at 0/1 samples, itās realtime.
It seems this fix is for the old hair shader (Chiang) from the original pbrt implementation it is based on, not for the new one (Huang).
Cheers.
Does this mean everyone has to build their own Blender version to work with ODIN2 for Cuda since redistribution is not allowed or I am wrong?
Thatās what it looks likeā¦
Dont we already have OIDN? What is happening someone explain it to me
We have it on CPU. What is being discussed above is OIDN2, basically a GPU-port (Nvidia only).
I need a video on this, build Blender from codeā¦ Iāve seen a video of OIDN2 running on a GPU, itās so fast and smooth! almost real-time unlike the Optix denoiser.
Hmmā¦Can you find anything about the license saying this? I found this: https://www.openimagedenoise.org/index.html
It is is released under the permissive [Apache 2.0 license]
https://www.apache.org/licenses/LICENSE-2.0
I canāt find anything about not redistributing.
I have no idea how to go about doing this tbh. Maybe the guy working on this can find some sort of way to do this or make a video on how we can add it ourselves?
Thatās pretty cool. Any chance you can share the video?
For the life of me I canāt find it, it was on Intel youtube channel though.
edit: found the video but it appears to be a number of different technology working together, GPU denoising is one of them. https://www.youtube.com/watch?v=WUFXJ2JtkNY&t=252s&ab_channel=IntelNewsroom