Isotropix Forums

Aces

General Discussion about Isotropix and CG related topics

Re: Aces

Unread postby dboude » Thu Jun 25, 2020 12:49 pm

kirkr22 wrote:As of re-scaling overblown highlights into visible range. Isn't it kind of a post process that could be working with any color gamut?


By rescaling the highlights at the render level, you allow using brighter lights that give more energy to the bounces without burning your image. This can not be done in post-process. (except by recomposing the beauty but in this case, you won't have the result you want during the lighting process)

My 2 cents

Cheers ;)
Démian
Isotropix
Technical Artist - Clarisse Specialist
User avatar
dboude
 
Posts: 1509
Joined: Mon Jul 03, 2017 10:51 am

Re: Aces

Unread postby isoyann2 » Thu Jun 25, 2020 7:17 pm

So if the gamut would include ultraviolet we would get it bouncing too but would it be changing 99% of earthly colors that are falling in sRGB range usually?

UV is invisible, so after bouncing on regular materials, it should still be in the UV spectrum, and invisible, so, no point to include it in the gamut (which, btw, is based on a visible light model, usually CIE XYZ 1931), with the exception of specific materials like fluorescent powder that shift the uv wavelenght to visible light. Such behavior require a spectral renderer, which clarisse don't provide natively, but it is easy to cheat this kind of fx. Spectral rendering is heavier than RGB... it is always a balance between costs and benefits: like caustics, it really depends of the current production you are working on.. usually it is overkill, but If you work on a scene 100% light by UV with dancers wearing fluo makeup and glass prisms spreading the light in spectrums everywhere, a spectral renderer would be handy :) For most cases you don't want to pay for the overhead introduced by the spectral calculations.

I think Aces should help with rendering correctly something super vivid colored , something nuke neon giving proper bouncing of such. But for regular earthly things how could wide gamut change anything?


For regular things, the change is very subtle, often not noticable. But once you get used to ACES, managing ACES or rec709 gamut on a per project basis introduce more complexity, while simply working always in ACES is more consistent (you just have to change the ODT, usually sRGB or P3, depending on the target output). Another benefit (if you work on a p3 or wider gamut monitor) is to be able to design your film with the color you like (you know, this lovely blue/green color you see often in theaters and you can't reproduce in the small sRGB space.. same for pink but I prefer the blue/green :) )then, even if you are forced to clamp the colors down for the today TVs, you'll be able to output in P3, respecting your original design.


A grass that is perfectly within sRGB, how could it make something bouncing of it turns ultraviolet for example? Except if light itself is 90% ultraviolet.

When green bounce over green, then over green etc... , the more it bounce, the more it is saturated. So, it will quickly be out of the sRGB gamut (especially the green).

As of re-scaling overblown highlights into visible range. Isn't it kind of a post process that could be working with any color gamut?

Yes. ACES is packaged with a nice "S Curve" compressing the dynamic range in a pleasant way, close to a standard filmic response curve... having everything in the same system simplify things, but if you only want to deal with the highlight compression, it is way simplier to just use a 1D lut on your clarisse viewer.

Like "Filmic look transform" in Blender. It "fixes" highlights for both ACES or sRGB renders.

Don't know if they provide the filic look transform as a 1D lut.. hope so, because you are not supposed to burn this transform in your render, and you need to be able to do in in comp, outside of blender. ACES make this easy, by making all the color management app independant (like ICC... pity they never extended it to support HDR and other film industry needed features).[/quote]


And does it without any reddish shift I see in ACES renders

This is not related to ACES.. don't know which error was made but using a wider gamut is not supposed to shift the colors. Maybe the screen is a wide gamut one, and the ODT is sRGB? Or sRGB textures or colors not converted in ACEScg? or converted with the wrong IDT? ACES is nicely designed, but imho, the terminology could sometime be better.. for example, it is hard to guess (well, this is why documentation is important.. ) that the ODT named "sRGB" is not a 2.2 gamma with a gamut compression to fit the rec709, but a S-curve highlight compression + gamma 2.2 + gamut compression... lot of potential errors if you dont read, at least, the simplified documentation chris brejon kindly compiled.
isoyann2
 
Posts: 7
Joined: Thu May 09, 2019 3:28 pm

Re: Aces

Unread postby kirkr22 » Sun Jun 28, 2020 12:32 pm

isoyann2 wrote:
When green bounce over green, then over green etc... , the more it bounce, the more it is saturated. So, it will quickly be out of the sRGB gamut (especially the green).


Are you sure? I am not challenging this assertion but it's kind of surprising me . My physics knowledge is a bit foggy but I think diffuse color bouncing is light scattering like Reylegh scattering in the atmosphere where light bouncing of air particles many many times. If it would be working like you suggesting we would be getting only pure blue light like a laser one of a single super clean blue wave length reaching the ground surface and see no other colors at all.

Do you have any links about this "more saturated after each bounce" principle?

I'm trying to persuade our software engineer to make sky color more saturated blue but he says his dynamic sky represent physically based sky color and it's way within sRGB range visually.

My guess the grass should scatter all the same range of wave length (i.e. colors ) it does in its first bounce either. Just with less energy (wave amplitude)
kirkr22
 
Posts: 106
Joined: Sat Apr 11, 2015 3:27 am

Re: Aces

Unread postby isoyann2 » Sun Jun 28, 2020 7:07 pm

You can test in clarisse (or arnold, renderman, whatever...), use a card with a low saturation green reflecting (sharp reflection or glossy, doesn't matters) multiple time a white emitter, pick the color, and look for the saturation: it will be a little more saturated after each bounce (but also darker.. there is energy absorbtion with each bounce.. so you don't get a pure bright color with near infinite bounces). Of course, the reflection surface must be tinted.

Another very simple test. A Sphere on a plane. Same material for both: diffuse green, 0.7 saturation, lighted by a white envlight and a white distant. Now, pick the color of the top of the sphere.. 0.7 saturation, not surprising.. now let's pick the middle.. mmh.. 0.74, this is more than my albedo! Pick the bottom, between the sphere and the card, 0.85!
Do the same with 1.0 saturation on the albedo and pick again the same 3 spots, everything will be clamped to 1.0 saturation, which is not a problem if 1.0 saturation represent the limit of your eye, but we all know it is not the case at all in sRGB.
isoyann2
 
Posts: 7
Joined: Thu May 09, 2019 3:28 pm

Re: Aces

Unread postby esmith » Tue Jun 30, 2020 1:16 am

dboude wrote:Hi,

Nothing changes except the color space

ACES
Moutain_ACES.png


sRGB
Moutain_sRGB.png


Cheers


This is a fascinating conversation, I try Aces on and off almost constantly. And almost never, have I seen an image in my 3d view I prefer. I also notice this shift toward red, which on my red car in my latest render pushed some areas of highlight into a weird magenta. I want to use it, my brain is almost 1/2 melted each time I do. Perhaps I need to take an Aces rendered image into Photoshop for post process to really understand the value in using it's spec....

But as someone said, since 9 times out of 10 an image is being seen on a phone or laptop, at least in my case......I'm not sure there is a real image benefit for my required end result.
User avatar
esmith
 
Posts: 180
Joined: Mon May 22, 2017 4:11 pm

Re: Aces

Unread postby kirkr22 » Tue Jun 30, 2020 11:57 am

isoyann2 wrote:
Another very simple test. A Sphere on a plane. Same material for both: diffuse green, 0.7 saturation, lighted by a white envlight and a white distant. Now, pick the color of the top of the sphere.. 0.7 saturation, not surprising.. now let's pick the middle.. mmh.. 0.74, this is more than my albedo! Pick the bottom, between the sphere and the card, 0.85!
Do the same with 1.0 saturation on the albedo and pick again the same 3 spots, everything will be clamped to 1.0 saturation, which is not a problem if 1.0 saturation represent the limit of your eye, but we all know it is not the case at all in sRGB.


Thanks Isoyann, I see what you mean. Not sure I understand physics behind it but definitely it should be based on real thing somehow. Like some kind of color "filtering" chopping off some part of wave length range after each bounce. Still the question is if such chopping is linear in each new bounce of fading to non together with energy and whether bouncing would drive any real life matter colors like sky,grass etc beyond sRGB range quickly?


I recall the issue with grass colors that was huge decade ago when expensive TV and pc monitors started to make greens and reds way beyond sRGB gamut resulting in somewhat neon, acid grass color on any football match translation and same in games too. The signal is still sRGB and monitors explode it into nukes , red faces on photos , crazy greens etc. We had to make kind of sepia or indigo filters on everything to cope with it.
Last edited by kirkr22 on Tue Jun 30, 2020 12:30 pm, edited 2 times in total.
kirkr22
 
Posts: 106
Joined: Sat Apr 11, 2015 3:27 am

Re: Aces

Unread postby kirkr22 » Tue Jun 30, 2020 12:19 pm

esmith wrote: I also notice this shift toward red, which on my red car in my latest render pushed some areas of highlight into a weird magenta. I want to use it, my brain is almost 1/2 melted each time I do. Perhaps I need to take an Aces rendered image into Photoshop for post process to really understand the value in using it's spec....

But as someone said, since 9 times out of 10 an image is being seen on a phone or laptop, at least in my case......I'm not sure there is a real image benefit for my required end result.


I have just read through Chris Brejon paper https://chrisbrejon.com/cg-cinematograp ... imitations
onr more time and in the end , in 'ACES limitation' section : "It's responsibility of the colorist to manage this kind of problem by de-saturating reds a bit" mentioning ACES is not ICC color management .
kirkr22
 
Posts: 106
Joined: Sat Apr 11, 2015 3:27 am

Previous

Return to General Discussion
cron