Maxwell Render | Official public forum

Hi,

NL developers, please check this preview of V-Ray 1.5:

http://www.vray.info/features/vray1.5_preview/

The most cool part for me is the dirt shader. Please please include an accurate scientific dirt calculation method in Maxwell, if possible then preferably viewable in the viewport, like QuickDirt.

The best dirt algorithm I've seen so far remains REM Infografica's DirtyReyes:



So keep it Spanish!

Thanks and cheers,

Metin

NL has mentioned procedural textures will be possible in the future. A dirt shader should be one of the first implemented.

A dirt and grime shader would indeed be good and useful, but it's tricky to use procedurals for those. A common Ambient Occlusion shader with a few extra parameters for modification and tweaking to suit the dirt material to your object. That would do the trick and quite easy and render fast aswell. And you can also ofcourse bake ambient occlusion data on your mesh for even quicker rendering.

But dirtmaps are also based on a adapted type of algorithms and not really procedural noisemaps unless you add that as another variable to the equation to give the dirtmap some irregularity based on a procedural noise.

/ Max

"dirtmap" - "ambient occlusion" - same thing, same principle for working. For a dirt shader some noise would be useful.

A dirt shader seems to be a noisy ambient occlusion rendering. I would use a fast biased renderer for this.
I think, it is nothing for maxwell, because it is not am effect of the real world, it's a occlusion fake.

What you could try is, render your model in a grey environment. Could be done in Maxwell fast too and should looks like a occlusion rendering. Than you can combine this image in Photoshop with the scene rendering of environment you like.

Also you could try to denoise a poor Maxwell "occlusion" rendering. Maybe you get this noisy dirt look at this way. If you use a grey model, than you get a black&white image. This could be combine with any color image.

A first test without any caustic layer show me a very fast noisy SL 9 occlusion image.

PS: .. best you use a strong gamma 1... 1.4 and a high burn 0.

Thanks for your comments, but I don't completely agree with you guys. A correct scientific dirt algorithm may have a lot of visual similarities with ambient occlusion, but ambient occlusion is simulation of soft shadows caused by environment light, while dirt grows between each little concave polygon detail and has an important random factor and grainy noise factor. Dirt is caused by weathering, not by light, and it involves more fuzzy logic than the quite predictable ambient occlusion. Furthermore dirt depends on a number of other factors, such as the orientation of a mesh, and dirt grows on object edges as well, while those edges are lit by ambient occlusion because they don't block light.

Just my $0.02. I still hope that NL will incude a good scientifically correct dirt shader. And of course anyone's free to use it as a quick ambient occlusion solution.

Check out the DirtyReyes gallery for nice examples:

http://www.reyes-infografica.com/plugins/dirty_p.php

Cheerio,

Metin

MetinSeven_com wrote: A correct scientific dirt algorithm may have a lot of visual similarities with ambient occlusion, but ambient occlusion is simulation of soft shadows caused by environment light, while dirt grows between each little concave polygon detail and has an important random factor and grainy noise factor.

Metin, that's why I said a dirt shader should have some noise functions so it doesn't look smooth, but you will find a "dirt" shader without a noise function will work exactly like an ambient occlusion shader. They work on the same principle:

http://www-viz.tamu.edu/students/bmoyer/617/ambocc/

I hope it will be possible for us to write ourselves these kinds of procedural shaders in the future with the material editor.

... here a fast (2h) Maxwell Occlusion test. I think, if we got a noise function for the diffuse surfaces, than it would look a little bit more like dirt. It's a pity that I can not use a noise map in box projection in Rhino.

Micha wrote: ... here a fast (2h) Maxwell Occlusion test.

well yes, a white diffuse + skydome + indirect layer turned off may work as an AO pass for an exterior (well, sort of), but maxwell would need some kind of auto cutoff for every ray traced in order to get a AO pass of an interior.

Yes, in AIR it is called "max hit distance". I fear, this is real biased stuff.

Looking good.

Anyway, a material-based dirt method would enable nice things like rust on metal, so I still really hope NL will include it in the new material editor.

Cheerio,

Metin

Micha wrote: I fear, this is real biased stuff.

you bet. in fact i really doubt we will ever see something like that.

and I don't really know if we need that...

Guys, I think a scientifically correct dirt calculation algorithm is not biased, while an ambient occlusion shader is indeed a biased trick.

I think a dirt algorithm would be very useful to create convincing materials with a weathered look, without the need to spend a lot of time UV mapping all your scene objects and create your own weathered textures.

Besides, it is nearly impossible and not scientifically correct to hand-paint all weathered little dents and bumps, so that approach to create dirt, rust and weathering would not only take much more time, but would be more biased (allow me to use your favourite word once again) as well.

Almost nothing in this world is purely clean, so a dirt shader that emphasizes dust and/or rust in dents would only enhance the realism we all want to achieve. I can't think of a good argument against having a dirt calculation method as an option.

It's only a suggestion guys, no need to have a whole discussion about it with a pro and a contra side, unless you're convinced that my suggestions deserve all this attention.

Cheerio,

Metin

Metin, I would like to see a shader for dirty look, a shader more than an occlussion shader.