GI Rendering vs. No GI Rendering
JoeMamma2000
Posts: 2,615
There is a lot of discussion in the CG community over the awesomeness of GI (global illumination) and related types of rendering engines. These include unbiased renderers, physical based renderers, GI renderers, and all of the cool "real time" renderers that fairly accurately simulate bounced light like in the real world.
And what that means is this:
Let's say you're walking down a city sidewalk, and it's 1pm, and there's a clear, sunny sky. There's a bright sunlight hitting the ground and buildings around you. For every surface that sunlight hits, it bounces off that surface, inherits some of that surface's color, and leaves behind shadows. Some of that light bounces off into space, some of it bounces off and lights up nearby buildings, but some of it bounces off and into your eye. And that's how you "see" those objects around you...the light bounces off that object, inherits some of the color of that object, and bounces directly into your eye.
Everything you "see" is actually just light bouncing off that object and into your eyes
However, some of that direct sunlight hits a surface, then bounces and hits another surface, and THEN bounces into your eye. And sometimes it makes multiple bounces before it hits your eye. And each bounce the intensity of the light diminishes, makes weaker and weaker and more diffuse shadows, and also inherits the color of each surface it's bouncing off.
What the "old" renderers (like Carrara's basic "Photorealistic" renderer) do is consider ONLY those light rays that are "one bounce and in". In other words, all you see in the render is light directly from a light source that bounces ONCE to illuminate the object and then into your eyes. ALL OTHER BOUNCES ARE IGNORED. So if the sunlight hits the green grass near you, and bounces off and hits the building you're walking past and gives it a faint green hue and very subtle shadows, those bounces are totally ignored by the standard renderer. They just render the sunlight hitting the green grass and into your eyes.
So you are missing all of the secondary light bounces, which include all the soft shadows they generate, and all the different colors they inherit from each surface they hit.
On the other hand, the GI-type renderers don't stop at "one bounce and in". They calculate multiple bounces, and take into account the different colors the light inherits as it hits each surface, and also the increasingly diffuse shadows as the bounced light hits uneven surfaces.
So, GI is a good thing...especially if your concern is reproducing reality as in a photograph. Now, is there a downside to each method? Of course.
With non-GI renderers, unless you take extra steps, your rendered image probably won't look like a photograph. But it will render relatively quickly. On the other hand, if you understand how to simulate bouncing light when you light your scene, it allows you to generate non-photorealistic renders which might actually appear realistic. It allows you the flexibility to use your artistic skills to "improve upon" reality.
With GI renderers, you're rendered image will probably take much longer, but it will look like a photograph. However, your image will be, well, just like a photograph. Even if you don't want it to be exactly like a photograph. As with a real life photographer in a studio, you can add or modify lights and colors to achieve the image you want, but you are limited by real world constraints. When you add a real light to a scene IT WILL BOUNCE, whether you want it to or not. You basically set up your scene, hit Render, and you get whatever the renderer tells you you get.
Now, if you want the best of both worlds, a skilled artist can manipulate his lighting so that he achieves something that might appear to be 80% or 90% or even 100% like a photograph, while benefitting from the high speed and flexibility of a non-GI render. You can add accent lights that don't behave like real lights, and give you a special effect or highlight you need to help tell your story.
And all that requires is some basic knowledge about how to simulate light bouncing off of surfaces, inheriting colors, and leaving behind shadows based upon what type of surface it's bouncing off of.
Attached are two images whose only difference is this:
Image 1 was rendered with only a sunlight/Realistic Sky and using the GI renderer. It rendered in 20 minutes.
Image 2 was the exact same scene, done using the basic Carrara photorealistic renderer.....a sunlight/Realistic sky, and two bounce lights (spot lights). One bounce light simulated the sunlight bouncing off the beige tile floor and illuminating the building and generating diffuse shadows, and the second light simulate blue skylight illuminating the building and generating very diffuse shadows. It rendered in 3 minutes.
Now, could they both pass for "realistic"? Yeah, maybe. But the Non-GI render allowed me to highlight the front facing building a bit more than with a straight "you get what you get" GI render. And it allowed me to light the stone tiles on the building surface so they showed more of their natural beige color, as opposed to the washed out blue skylight you get from the GI render.
Is there a right way and a wrong way? Absolutely not. There are two different methods with different results and different efficiencies.
My only point is that it's probably a good idea to explore all the techniques out there to see which is best for what you need.
Daz 3D is part of 
Connect
DAZ Productions, Inc.
7533 S Center View Ct #4664
West Jordan, UT 84084
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2024 Daz Productions Inc. All Rights Reserved.
Comments
Do not know what kind of GI your using (fake,, or using and HDRi), although the bottom image has more realism, with lighting on pillar. Top image has a light (like a sun or what ever) reflection on window. Angles of lighting are off with light (sun), in window reflection.
Please ignore that....it's a flare from the camera-mounted fill light. I forgot to kill its reflection in the window. I suppose I should re-render it.....
EDIT: Okay, now it's fixed...but in the process somehow I swapped the images.....anyway, now it's correct.
1st image is more realistic. Image 2 has lighting at bottom, moving up. Can tell by the shadows, and light balance. What is your objective here?
Thanks joe, another thoughtful and well organised post.
Robo2010, I think Joe Mamma is explaining, amongst other things, that you can mimic GI in Carrara and have better control if you fake it yourself.
As you probably know GI has render hit times in Carrara and we life is too short as it is without waiting for renders to finish ;)
Me, never been patient enough for GI ...
What is difficult (but not impossible) to fake is color bleeding.
If you look at the top of the central wall, you'll see that the roof color is bleeding on the top of the wall in the GI render and not in the non GI.
Otherwise, I completely agree with Joe. Not using GI allows more control over what is needed to be highlighted or not.
One other thing to remember is to use ambient occlusion. It helps a lot without the need for patience.
And as Joe knows but didn't say, you can also reference a real photo reference when you are setting up your lights, especially if you want to see how color bleeding really looks. I remember that from dabbling in a Vue forum years ago. Some really nice realistic renders were done by a couple people just paying attention to placing lights and negative lights in a scene.
There are a couple nice YouTube tutes for faking GI in Carrara using distant lights:
http://www.youtube.com/watch?v=K_m95FqGM5A
http://www.youtube.com/watch?v=pR45MP66i5s
Thanks for the info Joe.
There is lots of exploring, as Carrara has plenty of power for photorealistic rendering.
----
One way I use Indirect Lighting, is to save an Irradiance map(Pro only), which is basically a light map from the render. When you
render again with the saved map, it should take a fraction of the time. Mainly used for animations where only the camera is
moving, but you can also use it to adjust textures and shaders after, maybe for subtle changes to color saturations etc.
joe also knows that light, in the real world, cannot bounce. It reflects. I didn't mean to say that to give joe a headache! LOL
I am only saying that to help explain something. All surfaces in the real world have what is called reflectivity. The mirrors on my telescope have slightly over 93% reflectivity, which is quite good. Many household mirrors will be slightly less than that. Some really expensive ones may beat that. I could certainly have more expensive coatings applied to increase the reflectivity on my mirrors. But I won't because my eyes would likely not 'see' the difference, anyways.
Anyhow, clay that has been fired in a kiln, but not glazed will have a much lower reflectivity than an ordinary, smooth surfaced glass bowl. Etchings in either will further change their reflective values. Blades of grass have reflectivity and translucency.
As you can imagine, shape can do a number on how light reflects. In most video/image editing circles, you'll see the whole thing explained exactly as joe has, because bouncing is a good analogy to reflecting. But the difference in this case is that when you begin to analyze, even if incorrectly, in your mind the different reflective values of the surfaces within your scenes, you may have a decent starting point to estimate what colors are coming in from what angles, and by what degree. For example, a huge chrome ball between the camera and the subject would do what with the light? It would reflect a scattered array of all incoming colors - more so from directly opposite the light. Cast a bright light from behind the subject to reflect bright light back at the subject. This is a way that you can realistically have the source of light facing one direction - yet still illuminate the subject from a completely different direction, while still holding on to a certain believability, without any other apparent light sources. Keep in mind that higher reflection will have less color bleed, meaning less color from that surface gets reflected back. If the above example used a giant red balloon, your light would have a good deal of red in it, but would also allow some red light to come back toward the camera as well to maintain believability, due to the balloon's translucency. If it was a giant watermelon, much less light, slightly yellow/green would cast back to the subject without nearly enough intensity to cause a shadow. Just some quick ideas to think about while setting up your secondary(bounce) lights.
joe's wisdom in his top post really shines (pardon the intentional pun) when you see his example of using so few lights. Artificial Global Illumination rigs, set up to work for nearly any scene and situation within, will often use many lights. There is a good reason for that as well. More lights can give you smoother shadows without turning on Soft Shadows. But how much faster are they than using three lights with soft shadows?
Also, when it comes to getting your shot the way you want it to look, try turning the shadows off on some of the accent lighting you'll be using. Just cast shadows from the direction that gives you the results you want for the image. This is easily done by setting up that shadow-casting light first. Get those shadows where they will look best in your image. Now bring in light 2. tweak it to work. Now light 3, if necessary, and so on.
If your scene demands a certain believability, you can use the laws of reflectivity to help you out. If there's no way that light could possibly come from 'that' direction, put a reflecting object in to make it so. This doesn't mean you have to truly apply a reflective shader - don't take this whole thing the wrong way - just make it so that it is believable that the object could be the reason for the light.
BTW
joe, I agree with what your article states, and would personally chose the faster render.
All true...
But I think the other thing to keep in mind is your goals, and your audience.
There's a difference between "believing the premise" and "convincing photograph". You can give a 80% or 90% accurate render that will make your audience believe the premise, although you might not be giving a 100% accurate photo quality render. What you don't want is your audience to dismiss your render as fake and unbelievable IF your goal is to make a convincing 3D world. But on the other hand, do you really NEED a 100% solution? Maybe not. Maybe if you just apply some rough basic principles of lighting, and not go into too much detail and depth, it's enough for the average audience.
As has been stated, there are increasingly subtle color and light and shadow effects that occur with each additional bounce of light. So the artist must decide "is it enough to simulate ONE additional bounce, or TWO additional bounces, or...?"
Since the renderer only simulates "one bounce and in", that means you'll need to start with a bounce off a surface, then a second bounce off another surface, then into your eyes. Is that enough? Or should you go for three bounces?
The Non-GI render I posted was one additional bounce, and a very rough one at that. Often that's enough. But you have to keep in mind the following:
1. The color of the bounce light
2. The intensity of the bounce light
3. The shadows of the bounce light
So if you have a bright sunlight over your left shoulder (like in the image), you know it's gonna bounce off the cream/beige colored tiles and illuminate the face of the building. So you take a look at the tile surface, and decide "well, tiles are smooth, so they will provide relatively direct light with RELATIVELY harsh shadows. And the bounce light should have a slight bit of cream/beige color inherited from the tiles it's bouncing off. And the intensity should be RELATIVELY high since the tiles are light colored and shiny and fairly reflective".
So you set up a slightly beige/cream colored light sitting on the tile floor, point up at the building face, crank it's intensity down to maybe 20-30%, give it some soft shadows (but not too soft), and see how it looks. Often it's a balancing act of adjusting each or all of those three things to get it looking right. And as others have mentioned, there's nothing like a real life photograph, or just your daily observations, to see how light responds in the real world.
Myself been trying over the years avoiding GI, to bring in realism and faster renders. If I do use GI, I use only one distant lighting and render settings "GI" checking only "Indirect light" and under it, AO. If using, Sky Light, the renders are longer, and CPU intensive. Theirs been that "Fake GI", but the amount of lights to pull off realism is not easy, although the renders are very fast. If I truly want fast renders using GI, and will need $10 thousand to purchase 3DSMax, then I can make movies and such. So with my low income I have to settle with Carrara, although I do find expensive for the Pro.
3DS Max doesn't make lighting, or its principals any easier, better or different - and Carrara does an really good job of it, in my opinion.
And you can even get some fairly rapid results with Carrara's GI, even with full indirect lighting if you keep your scene fairly simple.
But the point is, you can really make nice renders without it - and without using a large amount of low level lights surrounding the scene - even though those can simplify if you're not that handy at figuring out light placements.
As for believability, I prefer not to think of such things, myself and would rather go for a particular effect that "I" want. Take the below image for example. I love the look of it. There is no possible way that such an amount of bright light could possibly make it through all of that hair to cast a highlight onto the top of her collar. But it really helps to sell that she is quite wet - and in relative darkness. I just have that light ignore her hair and duplicate the light and have the new one ignore everything but her hair.
You have to take caution with those types of setups, too... because it will have the hair cast no shadows from such an intense light. So I'll often have a third light that catches everything. BTW, this image was not meant to be realistic, as explained above:
Were chatting about realism renders avoiding GI. And regarding 3DSmas the program can pull off realism and not using CPU intensive. IF this guy can pull off a short movie "Code Guardian" using a Pentium 4 and using 3DSMax4 makes me wonder. The short film is not that realism, but http://www.cee-gee.net/ 3DSmax is way to steep for a price. http://www.youtube.com/watch?v=wjkcwA8Bk34 This short film was released in 2007, and what is it now? 2013, and we still can not pull this off, even in Carrara, no matter how much I keep trying.
That's well done, but he had help from the credits. Multiple Pentium 4s as well as he makes it plural in the credits (probably a render farm at one of the thank yous). And it wasn't just 3ds Max by itself that he used... looks like he used the very nice and pricey Dreamscape plugin for the ocean and clouds from http://www.afterworks.com/ . The fire and explosions looked like they were composited in. But he has a ton of skill and talent and he can't be a first timer as it's too well done. But we all keep trying. One of these days, someone will pull it off, maybe you! Remember, what Carrara can't easily do can be composited in with some good planning.
This is my realism (Carrara 8). All I could get. First image is GI (15min render time), no ambient, and last image is No GI (5min render time), ambient 25. Had to turn around character for lighting. Only one light for both images. Still needs work. Not happy with em. Still working on realism.
Animate her well, and she'll give you a very realistic motion picture. Something that can help speed you animation render time in whatever software you're using, would be to back off that camera some. That close up might be needed here and there... perhaps even a lot. She'll hold a great value for you if you smirk with her a witty comment or emote her way as she listens to someone talk... a great set of texture on that top quality mesh can do nearly anything you want.
As for 3DS Max, I don't think I'd agree with its pricing. I'm fairly certain that it's price is right on the mark. Every single friend or acquaintance who makes a living from it is really addicted to it in his or her spare time as well - and has a really talented full time staff creating its functionality. SMRP on that is a good deal for what it is capable of.
The truly capable Carrara simply has a very kind price to help get more people a chance to find their way into the art of rendering using high end techniques.
I did watch all the way through the thirteen minute YouTube yet - but what I saw so far is very cool. What I see is a very well planned series of clips that certainly could have been pulled off in Carrara. I know this because this is the sort of thing I do in Carrara. While my stuff is quite different, the efforts in getting where I am now has taught me that... yes. Carrara is a very capable 3d rendering suite. The artists who made the sample videos you've linked to could have said that they made this in Carrara, and some of us would not question it.
The hardest part is to have everything in the clip just right. Camera movement, acting quality, scene environment including atmosphere, movement, lighting, and detail (or lack of) all has to be working right - which is the ultimate challenge that will often pass through several hands, where each portion I list above is done by a separate department of one or more people to accomplish. That is why I don't feel so bad that it's already taken me years to only get to where I am now - which is at a point where I know that Carrara is truly capable of being responsible for all of the 3d modeling, animation, composite 3d with video, setting up atmosphere, lighting all in motion while providing the ambient moving somethings in the background as well as assisting big time with the special effects. Although Carrara is capable of using audio, I save that entire portion for an entirely different studio. Dogwaffle is excellent for many of the post work and even pre-work enhancements with it's full keyframe animation timeline completely independent of the input video - with many new feature to assist greatly for rotoscoping to add, remove, enhance, or otherwise change one thing through the entire clip.
Game engines are capable of doing some great action real time and can be scripted to record specific sequences controlling any number of actors to and from their own waypoints, which specify what the actor does upon reaching each waypoint. Home users have been able to hack into game engines to do this stuff for a long time, and later games even offer some sort of way in - if you have the utter ambition to try and do what their team did over the past eight years writing and getting just right.
It all depends upon your determination. Build a team of Carrara people who each like to do something different, and if you all groove nicely, it could turn out some very intriguing results that the individual user may never discover on his or her own. This is where the community and the Daz3d store has truly become my team. I have a long list of asset creators who work for me at such low rates that I was able to acquire years worth of assets on my meager income. The magic of the Daz3d co-op! Lighting, modeling, rendering, animation (etc.,) experts only cost me my time. Check it out... joe just inspired me to make some scenes using as few lights as possible - and only using soft shadows where absolutely necessary. That tip alone could save me hundreds of hours getting my super high def textures on my actors rendering closeups, pans across vast, gnarly and foggy terrain, a chase scene from throughout the confines of a metropolis, escaping to the tightly knit steep mountain cliffs of the roasting desert, all to end up under water, where our heroes must don their scuba gear to complete their mission of annihilating the saboteurs!
Carrara can render, my friend. It's right up there with all other render engine-running software options. The staff developing it have a job on their hands with this baby though, because it offers so much. It seems to be easier, or more workable(?) to add new features than it is to beef up what's already there, it seems - or perhaps they just need to know more details of where they should take the code - I don't know anything, well... not that much... about that sort of thing. But if you put your giant, Nazi Darth Vader-looking guy into a Carrara ocean primitive, and render out all of your motions, either with everything else in, or doing it in layers of 3d over video composites, you can use Project Dogwaffle: Howler to add and/or enhance things in between or after. Like the water draining hundreds of gallons out of his elbow as he lifts his arm back out of it, or the bright flash and flare effect - perhaps even some smoke, sparks or running fluids can either be automated, painted, or simply letting some presets between a couple or a few keyframes...
The more imagination that gets poured into a production - especially from various angles which all understand where the outcome should be, the more option you have to choose from during the final cuts. A few seconds at a time, and you'll work your way through as far as your inspiration will take you.
Carrara rocks with its abilities offered toward automated animation settings - like for fire and ocean.
Fire is not an easy effect to achieve realistically no matter what software you use. The best results I've seen use real video footage of fire. Some of them used layered composites of the footage, while others used part of clips of the footage launched through particle emitters. I found a really cool spot where Carrara's fire primitive looks great for what I need for one thing, but now I need to figure out how to make fire work iin another area, where I have been unsuccessful so far with the fire primitive - but not with Carrara - for Carrara can implement the other methods and many more as well.
Carrara is actually a lot more diverse than most others. But all of the higher end applications have held their own against their prices over the years - because that's how their system works. The Carrara method rely more on the artists wits to figure out how to solve some problems using the many tools that are already installed in Carrara right out of the box - where another more expensive package might have some sort of primitive for it. Blender has had a very accomplished team of developers for a good run of highly enriched enhancement over many years now - where it once was more of what you'd expect from something that was entirely free. Nowadays, that free package is trying to contend with everything that's out there.
Whoa... sorry for the long post...
I get stoked up on the abilities of Carrara - for they are many, and they are good. You might have to run a few passes through to get the effect you're looking for - but so do the other guys, so that's okay! :)