CoolColJ's test pics thread

[CoolColJ]

decided to finish up and render my old CG Sphere entry I started years ago

Quick test render - final has to be 800x800. Should I do it in bidirectional pathtracing or bidirectional MLT? MLT tends to have a smoother look, but there are no complex light paths or much caustics

Gonna take a while, it renders quite slowly....

[CoolColJ]

^ after doing a few full size test renders, it looks like Bidrectional MLT is the way to go. With Bidirectional Pathtracing, all those little holes and pathways in the model don't seem to be lighting and clearing up as well.

[CTZn]

no such crashes with 2.2 (aka 2.0 stable) indeed.

In fact you can even pick 2.4.2, the issue is specific to 2.4.3.

[CoolColJ]

no such crashes with 2.2 (aka 2.0 stable) indeed.

In fact you can even pick 2.4.2, the issue is specific to 2.4.3.

thanks - good idea


I have a fish for my pool
All native C4D materials, I just hit export and all the textures and alpha maps exported just fine
Those hot pixels decided to show up right before I stopped the render though...

[CoolColJ]

Alien sphere CGsphere entry has finished rendering
http://www.indigorenderer.com/forum/vie ... f=4&t=9727

[CoolColJ]

After much testing I find Indigo the slowest of the renderers I've tried when it comes to rendering underwater caustics, lit by the Sun

But under artifical lights, it is one of the fastest!
With Biddirectional MLT

same scene, all rendered for 35 mins -
Instead of patch by patch like in Lux, or a collection of dots, in Thea, Indigo does big broad, smooth strokes.Although I still don't like the multi-coloured artifacts

order - Thea, Indigo, Lux - normal Bidirectional MLT, Lux BiMLT with "use variance" option

Why can't Indigo be like this under the Sun?

[lycium]

The reason it can't be that fast from sunlight is that it is much more difficult to sample. We have some plans for addressing this, it's a little ways out though - this problem is one of the hardest in unbiased rendering.

As for the colour noise, that we can improve in the near term, like the next build

[lycium]

I'm just trying some improvements to reduce the colour noise now, can I get that scene for testing please?

[Zom-B]

After much testing I find Indigo the slowest of the renderers I've tried when it comes to rendering underwater caustics, lit by the Sun

If I remember correctly you are on C4D J ?!
In Cindigo per default Sun renders in 2 layers, sun + sky! Disable this in Advanced Indigo Settings for a speedup!

I'm just trying some improvements to reduce the color noise now...

This sounds great.... will this also affect SSS color noise?!

[lycium]

Not all SSS colour noise; if the scattering is wavelength-dependent there's no way around the colour noise, and MLT is an absolute must.

[Zom-B]

yes, SSS without MLT doesn't make any sense

I just realized in this renderings that the Sphere from Indigo has some different IOR compared to Lux and Thea!
Could you please check this J?

[CoolColJ]

lycium - Will do, I'll email you the scene
Will be nice if you can fix those multi-coloured stuff. That has been bothering me for the longest time in Indigo! Especially MLT without bidrection!!


Zom-B - Will have to try render without the 2 separate layers for Sun+Sky.
What is the purpose of this option?

The IOR of the sphere should be the same for all - 1.6
It's using a C4d material, and checking the scene file it is exported as 1.6

[CoolColJ]

BTW with MLT you can turn large mutation probability down to zero and Indigo will render just fine, as long as max change is between 0.02 and 0.015. Too small it becomes blotchy.

Infact the large mutations kinda make it more grainey....They're basicly random path tracing samples I presume
Funny thing is, samples per second drops the lower Large Mutation Probablity is. That tells me Indigo is working harder

very interesting

I've also been experimenting will really high Max Consecutive Rejections, like 100,000 to 10 million , and it really helps a lot in Lux render. Allowing it to render sun caustics within 30secs and have them well developed within a 100 samples. Not so much in Indigo

[lycium]

Hey presto, colour noise is suppressed

I changed your MLT settings a little:

Code: Select all

    <large_mutation_prob>0.35</large_mutation_prob>
    <max_change>0.0035</max_change>

[lycium]

BTW with MLT you can turn large mutation probability down to zero and Indigo will render just fine, as long as max change is between 0.02 and 0.015. Too small it becomes blotchy.

Infact the large mutations kinda make it more grainey....They're basicly random path tracing samples I presume
Funny thing is, samples per second drops the lower Large Mutation Probablity is. That tells me Indigo is working harder

Guideline for max change: if you have large areas of smooth illumination, maybe use 0.01, for lots of bright small lights use maybe 0.005, and consider not using MLT at all (except if you have SSS or something).

Large mutation probability: in scenes where the light transport is really difficult, use ~0.35, in scenes where it's simpler you can use maybe 0.25 to get more efficiency.

I've also been experimenting will really high Max Consecutive Rejections, like 100,000 to 10 million , and it really helps a lot in Lux render. Allowing it to render sun caustics within 30secs and have them well developed within a 100 samples. Not so much in Indigo

MNCR only matters in really difficult scenes where MLT gets stuck on small bright sources. Using a higher value actually makes this worse, whereas using a lower value will introduce bias into your image but allow MLT to explore more. Guideline: don't go below 500.

I've had some ideas recently on how to handle scenes with difficult specular light transport, it should be really effective (and isn't very complex). However, right now we're crazy busy getting GPU support out, and after that we need to do some solidifying and more GUI work to put out a proper stable release.