A big page of CGI realism tips that mostly is applicable to Blender

that is not particularly a good example of realism, it nice but not realistic. And the two biggest give aways are the shadows and the reflection; enviroment map and shadow buffers.

You could pass this off as realistic if this was an animation but for a still were people can stare and analyse these kind of things become obivious

OK, you’re right. Maybe if I put a little more work into it, but my aim was how far can I fake realism with a minimum of features. Just like Pixar did in the old days.

I agree totally with the article I don’t think that anyone is bashing BI.
My opinion is that nothing beats a good GI solution. But if the result is not satisfactory, the renderer should have enough tools for added artistic freedom (a good example is Harkyman’s colored shadows) so one can cheat his way into photo realism.
By the way… wasn’t there a semi fake caustics patch a few years ago for the BI?

I respectfuly disagree. The word photo-real in itself is a cheat. It implies: “How we are used to perceive a photographic image”. It has a lot to do with the behaviour of of a photographic camera. Tricks like lensflare, film grain, DOF and such ads to ‘realism’. A badly taken photo stills looks ‘photo-realistic’.

OK, but if we leave out the extremes, there is an area in the middle that we can safely say that such CGI is photo realistic, and it can be achieved with an average set of tools.

I was going to make a comment earlier and decided against it, but it dovetails with what toontje just said so…

I think we should be talking about believability, not photorealism. Believability simply means that it looks like it actually exists. IMHO, believability is often a better target than photorealism.

Right you are! That is the word what I was looking for too.

Well said, harkyman and toontje.

I particularly appreciate and endorse the point of believability being a more appropriate and useful goal than realism.

This is a much better strategy in the long term, in my experience, because it can also suit techniques not consciously striving for realism, as in NPR and toon-shaded renders: characters/objects/scenes can be convincing in their own stylistic contexts without necessarily having to conform to physical reality.

Blender is already phenomenally powerful. We can create relatively believable images internally even without raytracing or AO (one example being a render I just uploaded), and quickly so (this render took less than 30 seconds).

More tools are always welcome, of course, but, for me, nothing beats the speed and flexibility of Blender’s internal rendering capabilities :slight_smile:


Robertt, that’s a nice piece of work, and it illustrate exactly what it is all about. Photo realism is suprisingly a narrow field, mostly for archviz, hot looking babes etc. But there is also something I heard someone from Pixar said before: Hyper Realism. He didn’t elaborate on that, but I think that your image illustrate just that. Some kind of alternate believable reality.

One think I miss in that Cgtalk thread is that many if not most materials, related to human activity, are in fact a mix between several properties, and in many cases a juxtaposition of identical properties but with different incidence.

For instance many metals show several kind of reflections juxtaposed, each one related to the finishing or polishing treatment given to the surface. For instance, a metal can show a anisotropic reflection made by a rough longitudinal polishing and isotropic glossy reflection made by a final radial fine polishing, all at the same time.

Some other surfaces show different levels of reflection, one ‘original’ reflection as the object got off the factory, and a posterior ‘human touch’, an anisotropic reflection as the surface got eroded by human activity, all at the same time.

In fact, any kind of surface related to human activity has got some percentatge of glossy reflection on it, and materials with a very small percentatge of glossy still show fresnel effect in the macroscale.

That’s why pure materials should be always avoided, at least with artificial materials, and use options that let you mix and blend properties, and juxtapose the same propertie but with a different incidence in each level.

Excellent points, Alvaro.

I think it’s also important for users to remember that they have so many options when it comes to Blender, whether it is rendering externally or internally.

Internal rendering is an area I decided to specialize in since way back for numerous reasons, including using slow computers in the past.

Even now with a faster computer I am enjoying the flexibility of Blender internal more than ever.

In fact, this very thread has inspired me to expand upon the topics of believability and internal rendering, and I just posted a new thread in Blender Tests which also includes somewhat of a tutorial.

In getting beyond the perceived and physical limitations of the internal renderer while evading the punishing rendering times of external/unbiassed renderers, we can learn and introduce new techniques in our workflows that can bring us more quickly to the sense of believability.

The solutions and the end products might not be physically correct, but, total realism is not always the optimal goal for every single project. Often, going by my experience, less realistic approaches can produce more interesting and better looking results.


This is largley dependent on the kind of work you are into if like me you are a gear head like making cars than a physically correct render engine is a big help.

I choose cars here because this is what led Pixar to revamp Prman and add raytracing and led to the guys that did vfx on speed racer to write physcially correct shaders for mental ray.

Arch viz would be another field where I would think having physically correct shaders would be a help.

but physcially correct shaders are not only for unbiased render engines yafaray is biased and so is mental ray but you have physically correct shaders for both. In fact I have seen BDRF lambert shaders in Renderman shader language.

Indeed, which is why I wrote in the previous sentence and post:

Having all methods available would be ideal, it would seem.

However, until that ever happens, there is still so much we can learn to do more with the existing tools we have.

Rather than for any user to wait and bemoan the fact that Blender lacks a certain capability, I would prefer that hypothetical user to invest that energy more positively and find a way to create what wants to be created.

This is why Pixar and others were required to innovate when they reached a perceived or physical limitation.

This is also why we must individually be open and conducive to innovation, be it in tools or inventive techniques, and remain creatively engaged in all aspects of the project, including through the rendering process.


Yes, blender has a lot very useful tools and in many tasks its proper tool to use as a renderer, but when we are adding 3d to live footage, then you start struggle because you have reference where you need to make it believable (physically correct materials does not mean its needs to be unbiased renderer as Tyrant Monkey said. It does not even mean that you need to use GI).

Believability vs Realism. Try to convince and explain that to your client when you have done some 3d for him/her, who is already afaird of 3D because things usually looks so CG :). Clients are not blind, and do not underestimated them. I have done it, and learn from it :).

I have shared your opinions before I went to work in post production house where these kind of VFX work we do. My opinions has changed by then dramatically and for reason.

I think comparing things how and what pixar does is bit unrealistic in that way that they have huge staff who works just for materials etc. Realistic is to think what small group can do and how to do it efficiently, and for VFX shots its efficient and more friendlier todo in team things with physically correct materials than spending huge amount of time just to imitating it with artistic hacks. Artistic hacks are meant for final touch (yes, we also use compositors in our pipeline at work. Blender isn’t only solution which provides that possibility).

Its also more realible because physically correct materials are more predictable. When its realible and predictable, you dont have to render animation over and over again just to get materials behaving correctly. Also GI and physically correct materials may not necesserly mean longer render times, because I have witnessed couple times that rendering with GI in vray is faster than rendering without it ;). (yes, vray is fast, and may not compareble in this case).

and yes… I love blender and Im only stubborn blender user in our company. Thing is that I would love to make renderings with blender, but right now i can’t. Still I try to use blender whenever it is possible, I have strong faith on it :).

ps. this really isn’t feature request for blender. Just want to share my experience because i feel that this community has wrong believes what comes for physically correct route.

Thanks for the informative post, 3dZ.

The scenario you describe, in terms of compositing/live action/VFX, is often a challenge in any pipeline.

In my opinion, believability easily trumps realism, because it’s the thing that translates over all media and styles: it’s the thing that allows a few well placed pencil / pen / brush strokes to suggest a character, the thing that makes us look past water color or oil paint or metal sculptures to see and connect with the underlying concept and emotion.

Realistic glass or chrome on its own is not that interesting or aesthetically pleasing until it has reflections, and people are more likely these days to rely on HDRI or environmental maps rather than bother to model a complete environment to facilitate such reflections - as in many car renders I see (especially ones with the fake building reflections when they’re parked in an open road or field - I just have to smile at those).

Even realistic photos or live footage itself often needs some post-processing to become more usable, interesting, and consistent. Composited CG elements, tracking, light probes: these are all areas that can involve significant work, and materials/shaders can be much less of a problem than motion:

E.g.: a poorly tracked / animated / composited / lit model or poorly timed / rendered performance/motion-capture will result in a much more noticeably inferior product much faster than if the object or character’s shaders’s not realistic. In fact, in live footage and animation, you often have a little more margin for something less realistic because it and/or the camera is in motion.

Considering it the other way around, it’s the overall relative believability of a performance that allows us, for a moment, to believe a character is alive and/or that a situation is truly unfolding (in the context of the work’s own subset of reality).

Shape and motion are high priorities among viewers: if something looks off or doesn’t move naturally, this can justify criticism of CG elements if they are improperly handled.

Another example is how we see, all the time, realistic looking materials and textures on 3D characters and scenes but the scale of the textures or things like proportions can be off, if only just a little: it helps underscore how the image is CG.

None of this 3D stuff is truly real: most of the algorithms themselves are not perfect or universally workable or optimal in every scene imaginable, and, even the few that come as close as possible, do not tend to function well in real time or fast enough for animation purposes: learning new and alternative methods can become necessary at some point.

My own feeling is that Blender is just a tool, and, like any tool, it can be used in a variety of ways for a variety of tasks but cannot possibly be and do everything all the time.

My only real interest in this thread has to do with Blender’s often prematurely dismissed internal material/rendering capabilities. It can handle more than most people give it credit for. After thousands of projects big and small over the years in many different styles, I’m still enjoying seeing what else it can be made to do :slight_smile: I’ve yet to really explore material and texture nodes.

So, for me, Blender still has much unexhausted potential :slight_smile: