Cycles Development Updates

[marcatore]

@melvi

Another thing I wanted to recommend is Cycles devs looking into maybe adding a color bleed slider to control the amount of color gi bleeding

Theorically, if you use “correct” albedo values, the gi bleeding shouldn’t be so cartoony.

@Ace_Dragon

I think now is the time to finally finish up and polish the microdisplacement functionality in Cycles to encourage people to start using it more instead of bumpmaps (which will boost realism as you no longer fake the effect).

I hope too because it’s a tool that sometimes could save and solve some situation, avoiding to have bump/normal shading issues or go to sculpting.

[Hadriscus]

@melvi

Another thing I wanted to recommend is Cycles devs looking into maybe adding a color bleed slider to control the amount of color gi bleeding. I am not sure if that might be why Cycles has a very cartoony look compared to other realistic renderers out there.

Color bleeding is “just” diffuse reflection, not sure what could go wrong there?

@melvi

I was a bit perplexed when the realistic movie that Blender Studio was working on called Charge was rendered with Eevee instead of Cycles.
Is this because they found Cycles realistic look unsatisfactory or there were other reasons?

My understanding is they wanted to production-prove Eevee specifically

@melvi

I agree that this should be worth the 10% drop in performance. Corona and Luxcore are not as fast as Cycles but their bump/normal maps look good and interact with light in a realistic manner.

Yea absolutely. I don’t work with normal maps a lot, but seeing the comparison shots convinced me

[Kologe]

@Hadriscus

Color bleeding is “just” diffuse reflection, not sure what could go wrong there?

One thing which does go wrong here is how the colors (of indirect lighting) shift progressively towards the notorious six (red, green, blue, yellow, cyan, magenta) with every bounce, as a side-effect of rendering in RGB (rather than spectral).

E.g. start from RGB [255, 255, 8], suppose we loose 50% ‘light energy’ after each bounce, one bounce later its RGB [128, 128, 4], after two bounces RGB [64, 64, 2], after three bounces [16, 16, 0] and then the skewing starts.

greetings, Kologe

[kakapo]

hm… does cycles really use 8bit color components? i thought it uses floats?

[Kologe]

@kakapo

i thought it uses floats?

Well that’s true, but 0 as integer and 0.0 as float doesn’t make a difference for the blue channel in above (somewhat contrieved) example.

greetings, Kologe

[SterlingRoth]

0 =/= 0.00195

But your point does still stand, there is noticeable hue/saturation shifting at high bounce numbers.

[melvi]

@Kologe

One thing which does go wrong here is how the colors (of indirect lighting) shift progressively towards the notorious six (red, green, blue, yellow, cyan, magenta) with every bounce, as a side-effect of rendering in RGB (rather than spectral).

Thanks for the info. Any chance Cycles can have spectral as an option that can be turned off and on?

Seems there was a spectral Cycles Branch?;

Cycles Spectral Rendering

ONE MORE UPDATE: There is now a graphicall daily build available for Windows which will contain my latest progress in converting Cycles into a spectral engine. https://blender.community/c/graphicall/nkbbbc/ ANOTHER UPDATE: Over on this devtalk topic I’ve been working on modifying Cycles to be a spectral renderer from the ground up. Progress is slow since I’m unfamiliar with the Cycles code but it seems achievable. UPDATE: Since the start of this thread, significant progress and changes of t…

[SterlingRoth]

Integrating spectral rendering took some bigger architectural changes that the core devs were unwilling to support. The project is currently on hold. If you want to use the old builds you can play around with it, but spectral isn’t coming to the master branch any time in the foreseeable future.

[kram10321]

Thoughts on making Cycles into a spectral renderer

there’s also an extensive devtalk thread about it

[SterlingRoth]

you can also try to tame excessive color bleeding in your shaders, it’s hacky AF, but it can help:

[melvi]

@SterlingRoth

Integrating spectral rendering took some bigger architectural changes that the core devs were unwilling to support. The project is currently on hold. If you want to use the old builds you can play around with it, but spectral isn’t coming to the master branch any time in the foreseeable future.

I see. Thanks for the info. Found this but its 3.5: https://blender.community/c/graphicall/Cnbbbc/. Any chance there is a 4.0 build?

@kram10321

there’s also an extensive devtalk thread about it

Yeah…and its really long. Can’t find the time to read through it. It would have been nice if the op placed a link to the latest spectral build in the first post so people can easily find it but I think the build is still wip so maybe that’s why.

@SterlingRoth

you can also try to tame excessive color bleeding in your shaders, it’s hacky AF, but it can help:

Cool idea. Thanks a ton for sharing.

[lsscpp]

@melvi

I was a bit perplexed when the realistic movie that Blender Studio was working on called Charge was rendered with Eevee instead of Cycles.
Is this because they found Cycles realistic look unsatisfactory or there were other reasons?

Afaik the short main purpose was to showcase the old guy beard (aka GN) and the potential of Eevee rendering

[CarlG]

@melvi

color bleed slider to control the amount of color gi bleeding

For the very few cases this applies to, we can do a trick in the problematic shader itself; if diffuse bounces see this material, replace the ray with a diffuse shader with a slightly less saturated version of the color. Or go completely wild and change the color completely.

[TheRedWaxPolice (I know dead mesh when I see it..)]

Cool stuff…

WIP: Cycles: Add thin film iridescence to Principled BSDF

WIP: Cycles: Implement volume stack priority and nested IOR support

WIP: Cycles: Implement blue-noise dithered sampling

[Son]

Should be in buildbot soon? nested dielectric was so needed. Another dumb question is this going to be in 4.1 or 4.2?

[DNorman]

I would say 4.2, 4.1 is in Beta status - polishing what is already in. New features are added to 4.2 (which is Alpha).

All of these are very interesting, I look forward to them, especially nested dialectics, but they are all marked as “out-of-date with the base branch” so there is some catching up to do.

[TheRedWaxPolice (I know dead mesh when I see it..)]

@Son

Should be in buildbot soon?

It might take some time…
Those are just work in progress pull requests that will need to be reviewed and approved, so no ETA…
4.2 would be nice tho…

[AlexeyAdamitsky]

Dielectric priority is hugely important for solid product renders and of course other stuff. Love seeing those changes coming to Cycles. It’s getting so much better and closer to other production renderers it makes Cycles more and more attractive to me again.

Still there is a lot more to implement like fully featured adaptive render time adaptive subdivisions with camera culling option and also render time booleans.

[Metin_Seven]

@AlexeyAdamitsky

Dielectric priority is hugely important for solid product renders and of course other stuff. Love seeing those changes coming to Cycles.

I’ve always loved this in LuxCoreRender.

[AlexeyAdamitsky]

Honesty, whenever you need to render anything where dielectrics are intersecting like liquid in a glass, bubbles in a liquid or liquid in a bottle ect. the ability to control its priority is a must. One of the reasons I can’t use Cycles right now. It just doesn’t give you the good enough result.

Once you can see the difference, it’s impossible to unsee it. When dielectric priority is in it’ll be a solid jump for Blender in terms of realism. Well, considering people will be setting up shaders properly .