Altus denoiser

[3LP]

The default settings are progressive sampler at 100 subdivs (=10000 samples per pixel). I've never expected anyone to wait it out... For the number of calculations that happened, it's amazing it took only 9h Maybe will reduce it to 24 subdivs. Best regards, Vlado

You're right, and I'm wrong, it was the default settings but with the adaptive sampler.

Sorry about that
Stan

[vlado]

Ah ok. I got Altus to filter an animation finally, I hope to post the results soon.

Best regards,
Vlado

[Paul Oblomov]

http://funkyimg.com/i/23Z1q.jpg
http://funkyimg.com/i/23Z1p.jpg
http://funkyimg.com/i/23Z1o.jpg
Here is my results. I've reinstaled my system, to make altus work Now I know, that there's a problem with naming, when altus crashes.
Two pics, 5minutes each + altus result.

[^Lele^]

Quite the pretty job!

[WilleViz]

Two pics, 5minutes each + altus result.

That looks very promising indeed.

[jenya.andersson]

Ah ok. I got Altus to filter an animation finally, I hope to post the results soon.

Vlado

Great Vlado looking forward for animation results! Can we test with motion blur?
Anyone know pricing of Altus?
thanks

[^Lele^]

It reads and supports the animation settings in the common tab of the Render Dialog.
Right now it denoises as soon as the two render passes are done, which is ideal if you use GPU, less so if it's CPU-bound.

Let me know of any bugs. (p.s.: it WILL change your split channels render path in the vray vfb tab, while it won't save the scene. Be wise with your file versions.)

altusMax_v006.zip

[vlado]

I don't think it makes sense to filter the frames one by one, they will probably flicker too much. The denoising algorithm works best if it can see all frames at once.

Best regards,
Vlado

[Mnietek2707]

Something strange. The same scene - different perspective views.
One result from altus is blurred and jagged, another is crisp.
Why?

[vlado]

Something strange. The same scene - different perspective views.
One result from altus is blurred and jagged, another is crisp.
Why?
[ATTACH=CONFIG]26733[/ATTACH] [ATTACH=CONFIG]26734[/ATTACH]

Can you post the input image files somewhere? I had issues when using the point position pass; things generally worked better without it.

Best regards,
Vlado

[Mnietek2707]

Can you post the input image files somewhere?

Sure:
http://we.tl/Em5HwGSZnk

[Mnietek2707]

things generally worked better without it.

True - it's OK without point position pass.

[Paul Oblomov]

Another try. 20 minutes per seed + 5 minutes to denoise. Full size png here http://rghost.ru/8s9PQcb9r

[super gnu]

hm.. im not sure why but so far, particularly the interior stuff has not blown me away.. it all looks a bit smooshy and rather like an imap with too much filtering and not enough subdivs. , even if, like paul's last example, the source images are not super noisy. are there are any settings to tweak? cos some of the other samples, Chaos' ones in particular, seem more impressive and less "filtered" maybe its just interiors being a pig as usual.

[super gnu]

out of interest, do the two seed images need to be equal in noise level? i guess so.

if one could be a longer render, you could use one nearly clean render, then altus + a quick dirty render, to remove the last traces of noise.

having to do 2 matching renders makes it impractical for doing a quick post cleanup on an image with traces of stubborn noise in.