Altus denoiser

[^Lele^]

Max to Altus SAMPLE script (and a few images...)

image 1: 2 x 30second renders (like the one to the left), one 8 seconds Altus pass.
image 2: 2 x 1minute renders (like the one to the left), one 13 seconds Altus pass.
image 3: 2 x 1minute renders (like the one to the left), one 18 seconds Altus pass.
image 4: 2 x 17.5 minutes renders (AA of 16 rays), one 62seconds Altus pass.



the script is here: altusMax_v004.zip
It's a SAMPLE script, on how to go about it in max.
Feel free to modify it, rewrite it, make it generally better and more suited to your needs.

Simple instructions:
Open a scene, set the renderer up as by defaults, change rendertime to whatever you please.
Open the script, browse to a folder, and type a base name for the output (the script will generate automatically the two pass names for you.).
You're generally ready to press the big button at the bottom (say this out loud ten times in a row, quickly...).
The script will set the render elements up, change the output for you, change the MC seed, render pass one, repeat, and render pass 2, build the Altus command, save it in a batch file (for convenience), run the batch file, and save the filtered image in the output folder.
You can press the "O" button by the output path to open an explorer window where the files are.
The EXE ostensibly appends "___flt" to the image name, unfortunately.
There is no progress bar, let it run and don't fiddle with it. ^^

[chadstevens]

Thanks very much Lele, I will try it out soon.

Cheers.

[chadstevens]

Unfortunately I keep getting an error with Atlus in the CMD, I would love to know how to get this thing to work properly,

I don't think anything is wrong with Lele's Script it seems to do everything perfectly but after it runs Atlus doesn't seem to be working and gives me this message.

I have tested it with the "use GPU" unchecked and that seems to work, is this because of the Nvidia Cuda 7.5 toolkit?

I did notice that when installing the cuda Toolkit it said that I had no version of visual studio available and gave me a link to a few downloads, unfortunately I had no idea which one to download so I downloaded "community" is this correct or am I missing something entirely.

Cheers.

[^Lele^]

Ah, there, glad to hear that it wasn't the script.
Well, it crashes with GPU active for Vlado, it won't work for CPU mode for me, so take your pick :P

[chadstevens]

Ah, there, glad to hear that it wasn't the script.
Well, it crashes with GPU active for Vlado, it won't work for CPU mode for me, so take your pick :P

Oh okay thats fine then Thanks so much for the script, it works beautifully

Now we just need the actual software without the watermark. fingers crossed.

Cheers.

[^Lele^]

Oh okay thats fine then Thanks so much for the script, it works beautifully

Well, it's quite limited in its operations, but it can easily be expanded: the main variables are clearly named and declared at the beginning, comments are aplenty, and you all know where i live, in case. ^^

[3LP]

I'm quite bad at coding, so I would really appreciate if anyone could implement the fact of being able to send to BB, for stills and animations if possible.

Thanks would be very awesome!

Stan

[^Lele^]

why not Deadline too? ^^
And why not queue just the altus job, for a set of picked image pairs, instead of the render itself?
It's a VERY deep rabbit hole.
Jokes aside, it's very early days, and i doubt the coding effort would TODAY be totally worth it.
It's meant as a tester, for now.
Better to wait until the powers that be find a way to more tightly integrate the two tools together (I can dream of a VFB integration, if it didn't break Chaos' bank...), when they're both ready to do so.
This said, rendering a sequence, locally, in a simple, script-entered, start and end range is something i can quite easily add.
Unfortunately, or fortunately, it will denoise each frame before moving on to rendering the next (last i checked, it was too much trouble to change the code to denoise at the end.).

[Mokiki]

Is this the same tech that Renderman seems to have implemented in its latest or release? Or different company but same idea?

[vlado]

Is this the same tech that Renderman seems to have implemented in its latest or release? Or different company but same idea?

I think the principles are similar, but it's not the same code.

Best regards,
Vlado

[3LP]

All right guys, here is the HQ render.
render time : 9h10min
Version : latest nightly
settings : all defaults, just enabled the GI at default settings also.
This just make me think that Vray at default settings isn't usable in production, who will go for 9h render at 1080p?, anyways :



and the difference in PS between this render and the altus denoise :


You'll notice that we lost a bit of texture in the wall, timber and foliage, but overall, for a 2h render in stead of 9h, it's pretty good IMO.

Stan

[vlado]

settings : all defaults, just enabled the GI at default settings also.
This just make me think that Vray at default settings isn't usable in production, who will go for 9h render at 1080p?

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

[super gnu]

off topic, but can i humbly suggest a noise threshold is added to the default settings rather than a reduced max aa rate? ive never understood why anyone would ever use a noise thresh. of 0, so cannot figure why its the default, rather than say, the 0.005 default of the dmc sampler.

[vlado]

off topic, but can i humbly suggest a noise threshold is added to the default settings rather than a reduced max aa rate? ive never understood why anyone would ever use a noise thresh. of 0, so cannot figure why its the default, rather than say, the 0.005 default of the dmc sampler.

Yes, good point. The reason it's 0 is that it didn't work too well initially. We fixed it later, but the default stayed the same.

Best regards,
Vlado

[^Lele^]

The default settings are progressive sampler at 100 subdivs (=10000 samples per pixel).

With an MSR of 6 (default too) that's 60000 spp.
I bet Altus lost some detail in the high-frequency areas... ^^