Link is from Blendernation, but most of the stuff seen here can be done in 2.5 using the internal renderer if you leave out caustics and true GI. (realistic scenes using blurred reflections are now more possible to do with Jaguarandi’s work.)
I’m actually surprised that they didn’t have an example of gamma correction. And it is interesting to note that all the examples of bad CG they talk about at the very beginning was pretty much the only stuff possible when I started Blender at version 2.33 because arealight shadows and AO was real slow.
As for adhering to those rules, I personally think you can fudge it just a little if you’re going after artistic merit, but not if you’re strictly after realism.
Well… in blender fresnel isn’t based IOR at all, which also makes things really difficult. Also real specular reflections are almost impossible to make. I think that article just sum its up what blender really needs… Physically based materials (of course we can use other renderers than BI for these ), but with BI you cannot follow those rules well…
It’s more possible than it used to, when the first opensource release of Blender became available than for the average user making photoreal work he would’ve considered BI a piece of junk that even lacked things Anim8or’s renderer had.
Now they may be impressed with the BI related projects due to be worked on and implemented, and I say that before 2.4 BI wasn’t neccesarily usable for serious production work because it lacked render passes and other important features. Then came SSS, soft shadows for all lamps and blurry reflections/refractions, and now the volume renderer, and projects like the raytracer overhaul and true GI. Durian will surely add more things yet to the renderer.
You could get somewhat realistic renderings in a small number of cases but the realism level back then is equivalent to Bryce 5 or a well done render from Anim8or, you can see how much improvement to the renderer there was, and even then you can see in the gallery showcasing Blender in 2003 a number of examples of bad CG said on that site. As for realism, I have perfected a node setup which boosts realism for me which uses the image, and the shadow and AO passes to make a multiply map that darkens parts of an otherwise flat scene without darkening the areas recieving strong light from the primary lamps, the linear workflow option in 2.5 is a huge help as well.
Yeah, i have used blender since 2.36 version and i know current development situation and improvements. I was just commenting that its hard to follow those instructions if using blender shading system because blender shading system doesn’t follow real life rules. (I don’t say that we cannot get photoreal results, but its in pain in the ass trying to get material work like in real world.
quote from article which says quite a lot:
“Using physically correct route is a sign of an experienced cg-specialist. Doing things in the physically correct way (or as close to it as understanding of physics and maths allows) means that the results are predictable.
Layering artistic hack after artistic hack into shaders quickly results in setups that are unmanageable and hard to make changes to. Moreover, if you don’t light those materials in exactly the right way (as the original shader designer intended) the results can often be bizarre or just plain broken. This is especially important when you need to share shading and lighting setups between multiple artists working on different shots and sequences. Try to keep physical correctness for as long as possible and only branch off into ‘artistic licence’ when you absolutely have to. It makes sense studying photography and traditional lighting to know how to break the rules without breaking the physical rules.”
Layering artistic hack after artistic hack into shaders quickly results in setups that are unmanageable and hard to make changes to.
I would think that’s mainly with the artist’s shader setup like with nodes, but Brecht has said that the Blender shading system needs a top to bottom overhaul as even he agrees it is full of hacky code with hacks and bandaids, this may or may not be done for project Durian.
That’s the base of photo-realism. Without light bounces you can just forget about realistic renders. Nearly every pic from that site has GI or caustics used. So even leaving out that “most” it’s just a dozen of pics left. And taking into that “most” is not “all”, and area light in blender is not physically correct (it even doesn’t make the same result achieved with lamp-mapped plane), gives unbearable artifacts on certain light angles, all we have left are texturing and shader model pics. Just exactly like someone said: that article just sum its up what blender really needs.
There’re really great unbiased and biased nearly physically correct render engines over the internet being used in chain with Blender for years and BI could implement some techniques from them. At least photons and global illumination thing from Yafray would make Blender usable for a half of techniques presented in that paper.
::START DRAMATIC SARCASM::
Or we can just put all the Blender renderer developers in a lunatic asylum and BlenderArtists becomes MayaArtists, We can have a dartboard with Ton’s face on it and burn effigies of the Blender logo, you’re right, open source has absolutely no future at all, let’s put all open source wannabe professional app. developers in the asylum too and just make every Blender user buy MentalRay.
Or that’s what I’m appearing to get here, let’s also lobby for the removal of the Blender forum at CGTalk and put a filter on so on that site so Blender Internal becomes “Br Inl, it just sucks man, whoever uses it are jokes of the CG industry and should be in slums”, who cares how many stunning works in the gallery use BI. ::END DRAMATIC SARCASM::
Sorry for the dramatizing, but sums up what some people may think if they genuinely think the devs. wasted years of their time trying to make Blender’s renderer more than just the MSpaint of renderers, or the thoughts of a select group of Vray/Maxwell/MentalRay users.
NOTE: This is just dramatic sarcasm, I don’t think this way at all, I haven’t been sleeping well recently too, BTW: Arealights can now be gamma corrected in 2.5.
NOTE: This is just dramatic sarcasm, I don’t think this way at all, I haven’t been sleeping well recently too, BTW: Arealights can now be gamma corrected in 2.5.
Well, why the heck did you spend the time writing all that out then? To prove a non existent point?
As my point was, that blender has really good features and a lot has happend to renderer side and it still goes on, but currently it is almost impossible to follow those rules which was explained in article, because blender doesn’t have physically based materials. I didn’t say anything about that there’s not a stunning works here or blender sucks etc, only saying that currently when achieving photorealism to animations with BI, its pain in the ass to get base materials to work correctly in every light situation, just because those aren’t physically based. I dont want to spend my time too much to try get base materials behaving correctly. I want to spend that time to give my materials life/age/feeling with textures,… and then light it whatever mood I choose. That actual artistic job.
pretty much the same thought popped into my mind; if you use yafaray than things will be much easier and applicable. You will have G.I and more importantly physically based materials.
I usually find turning off the computer every once and a while gives me time to get some shut eye.
And I sleep on average about 4 hours a day for the past 1 1/2 years. Explains my insanity doesn’t it!
About physically based materials, isn’t that something the Devs are already getting planned to do. It in itself is a major undertaking and I think Brecht said it would take almost the same amount of time and resources as the 2.5 revamp to implement, but it is something they have on their to do list.
That said, I think you better get some sleep young CD. See on the forums:eyebrowlift:
I got a bit of sleep and now I don’t feel so groggy. It’s obvious I can’t spend all weekend staying up to 12:00.
Also, I turn 21 in less than a month, be lucky I won’t be posting while tired anddrunk (I don’t plan on drinking anyway).
About physically based materials, isn’t that something the Devs are already getting planned to do. It in itself is a major undertaking and I think Brecht said it would take almost the same amount of time and resources as the 2.5 revamp to implement, but it is something they have on their to do list.
The shading code must be a serious mess if it will take that long, that sounds like ripping all the shading code out and starting over, no wonder they’ve never recently accepted patches for new shading models.
Eh. I read this the other day and just went and re-read it. In some places, there are some actual tips for getting closer to believability in your renders. Most of it is just an extensive explanation of terms. It kind of drifts into an essay on “what would a renderer need to have to be always photorealistic?”
I can distill his actual advice in a small set that I’ve been following for years:
Don’t oversaturate your colors.
Pay attention to scale, as indicated by size of details like dirt, light source distance and the convergance/divergance of shadows.
UV Map your textures, including procedurals unless there is a specific reason not to (cleaved rock, etc.)
Fillet your edges
Watch the colors of your light sources
Use soft shadows, but pay attention to things like distance from light source, etc.
Always use Mist, Depth of Field (properly) and Motion Blur
From that standpoint, you can do all of this in Blender Internal.
AO isn’t GI, but it gets you a decent amount of the way there. Also, if you want to wait for it, you can throw a very low level, single (or double) bounce fuzzy ray reflection on everything in your scene.
I actually tried that doublebounce reflection for more or less anything… I also tried to similarize hardness, roughness (from oren nayar) and gloss… My current formula for this, based on hardness, would be:
hard^(1/(ln(511)))/e = gloss
and
(1-hard^(1/(log(511)))/e)*pi=rough
Though I’m kinda pretty sure, those formulae are wrong^^
It’s based on hardness being an exponent on top of a cos-function
Like: cos(xpi)^hardnessintensity - or something^^
And I assumed both roughness and gloss to be linear, mostly because I couldn’t find formulae for Oren Nayar and don’t exactly know how the gloss is done…
Anyways, the approach you suggested doesn’t work as expected yet (in 2.49) as this seems not to work fine with smoothed normals. They appear to mirror faceted…
Well, no
It lets you set hardness, roughness and gloss as different variables (which is nice for special effects)… I’d like to combine them physically accurate-ish ^^
Same thing AFAIK would work with IoR and Fresnel settings… They’re kind of dependend on each other…
Harkyman: “Good thing Blender does all that math…” in physically correct way (or as close to it as understanding of physics and maths allows) means that the results are predictable? I dont think so :).
toontje: I’m really sorry, but I can see that none of those isn’t used (for realism this is bad example. i.e there’s totally black shadows in reflective ground plane in light gray enviroment). Not saying its looking bad, but saying its not realistic at all. As material is looking good but going for another enviroment it may look totally different or out of box. That’s the point of having physically based materials, those behaviors are predictable, and once made correctly those are working in any light situtations as they should.
I want to point out that my intention is not meant to bash BI, just commenting how different workflow blender has vs article. I totally agree with article’s workflow when style is photoreal.
No need to proof that BI is capable of doing photoreal pictures or even animations, because I have seen some work which is!.
“You can even paint quite photorealistic with oil paints!”
