TechiWarehouse.Com


Top 3 Products & Services

1.
2.
3.

Dated: Aug. 12, 2004

Related Categories

Maya (3D Animation)

By Steven Stahlberg

Modeling Advice

It's like learning to speak a new language - a book will help you but there's no substitute for actually DOING it, and doing it, and doing it... For NURBS, I mostly use "Loft", occasionally "Create", "Extrude" and "Revolve", but I find "Loft" usually gives me more control.For poly modeling see my mini tutorial on organizing the poly tools.

You'll almost always have to edit the object's shape extensively after creating it, this is ok. In fact it's never perfect, there's always room for improvement. But knowing when to stop is something you learn with experience. A good rule of thumb - if you find yourself making excuses for the image to people who see it, even to yourself, then you should probably spend more time on it.

You might even have to rebuild the object - building a new one using the old one as a template. Don't be afraid to do that. It will always turn out better the second time. The thing to remember here is you need LOTS of patience.

Sketch before you start modeling.

Use whiteboard markers on your screen for sketching, marking etc.

Use the clipping planes and layers in your modeling windows to view narrow cross sections, or to reduce clutter, especially when the model starts to become complex.

Use a neutral-colored Blinn with a muted and wide highlight as default shader, to easier see small errors in your surface when modeling.

You shouldn't have to use "Artisan" on a face, if you do you probably have to many isoparms in it. Also be very careful with "smooth"; don't rely on it to fix everything because sometimes it doesn't (and again, if you have to use it you probably have to many isoparms in your surface, or an uneven parameterisation). You should be able to smooth any surface you create 'by hand' (a good exercise in any case).

"Artisan" is a good tool for creating certain kinds of surfaces, like some kinds of terrain (soft rolling hills etc), melting effects (wax, ice etc), gold nuggets, aliens, caves, some tree trunks, meteors, monsters, etc, but it's not really very useful for faces (or bodies), especially young faces.

A much better modeling tool is the wonderful Lattice deformer. This replaces the Proportional modification tool too.

You know when you've modeled something really complex, and you want to make a wide-reaching but small and smooth change to it? Perhaps you want to widen a head a little, without making the eyes oval. Pick all the surfaces you want to affect, press F8 and then select all the cv's or vertexes you want to adjust, plus a few more around the edges. Then add a lattice to this selection. Next go into the hypergraph, press F to center the new lattice in the window and then zoom out a little, there next to it you see the lattice base node that comes with it. Pick that also (Shift click). Now you can move the lattice, scaling it, rotating it into position etc, without the model being altered yet. Next you need to set the resolution of the lattice in all 3 directions so only the things that you want to move will move - do some thinking and planning here. If you don't want the eyeballs to move you'll need at least 2 lattice points in each direction "bracketing" them, this will hold them still.

Then you can go wild, and start editing the lattice to deform the head. When you're happy you can select all the surfaces affected by the lattice, and delete history, and the lattice will go away - though you can keep it too, don't delete any history if there are blend shapes or any other deformations applied on that face.

Another point - there's no need to finish all modeling tweaks before you start to apply morftargets to a face. Yes it's possible to delete history and edit the face and then re-attach the morftargets later, but an even easier and more flexible workflow is to put all your future tweaks in one or several morftargets - quick and easy, and you can always go back on an edit by pulling the slider back down to zero.

You can also use the pick walker function (on NURBS only) - move through the surface's cv's with the arrow-keys, translating as you go (press 'Alt' and the arrow keys become incremental translators of whatever is picked). Remember you can also pick and unpick single hull lines, very useful function. When deleting cv's pick a hull line (just one at a time) instead of single cv's.

Unless an edge or a corner is meant to be razor sharp or is too far away to be noticed always have some roundness or at least a bevel. It is also often more realistic, and shows the objects shape clearer by catching a highlight - see what I write below on edges, in "visual perception and composition", and in the shading section the 6 points for realistic cg.

 

error-file:tidyout.log error-file:tidyout.log

There are no real shortcuts, good modeling takes time, and what takes most of the time is the editing, the movement of cv's - the push-pull translating, this way or that, left one day, the next day or several days later you see it should have been right... But there's one thing that can make this process much easier, and that's...

References

Very Important: ALWAYS USE THE BEST REFERENCE YOU CAN FIND when you model (and shade, and animate)- and follow it faithfully.

Why? Because NOTHING BEATS REALITY for richness of detail, the organic balance between chaos and patterns, variations on a theme, and variations on the variations, etc. Even the best artists of all time spent their lifetimes studying the world around them, most of them using reality as their reference.

How to find the right reference? Think about what you're trying to do, sketch or visualize, then write a list of images that might help you (in the case of animation, a sequence of images). Then try to find out where you could possibly find these images... then go after them. I can't stress this enough - GO AFTER THEM - and USE THEM. If the image you're trying to create is important to you, getting these images must be just as important, because you need them to do it properly.

Some people don't really believe this, or they think it's just too much hard work (as do I sometimes). Some say they want to create the art that's inside them without the inhibiting effect of "visual crutches" or something like that - this is fine, if your goal is abstract or naivist art, although even such work can benefit from studying nature.

But if your goal is realism, you NEED good reference, you need it like you need your hardware and software. In the highly competitive world of professional illustration, many times it's down to: "The Artist With the Best Reference Wins". I know illustrators who will hire a photographer to do a full photo-session, just to get the perfect reference.

When you've found the images or objects you need, make sure they're not too small or too large to view comfortably next to your workstation. Now as a specific object you've got reference for starts to take shape on your screen, as it becomes more finished, concentrate, relax, flick your eyes back and forth between reference and model, forget what you think you know... measure relationships in the image ("that part's about twice as wide as that one, and two widths from that edge..."), this is all to try to get the right side of your brain working, the side which they used to say is the more 'artistic' and visual (but which has later been found to be an oversimplification).

One trick to help that along is to turn the reference image upside down, and the model as well. This helps to silence the the logical parts of the brain, which thinks it knows what something looks like without having to check it. But it doesn't, it is ALWAYS wrong when it comes to this kind of thing. You always have to check, and recheck...

In today's information overload it becomes an art form in itself to find the best reference you can.

Look at these two images - the left one shows a dolphin modeled in SoftImage by an expert, Mr. Saw Kok Hong, a talented Malaysian cg artist, but WITHOUT REFERENCE. The red lines indicate the changes made after looking at some reference images - the fins were all too thick and the wrong shape, the tailfin was too big, the eye too light, etc. (Though it's likely your mistakes would've been different ones.)

 

error-file:tidyout.log error-file:tidyout.log

This image shows the final result.

 

error-file:tidyout.log

Photo realistic modeling must always be an iterative process, where you go back and forth many times from your model to the reference, comparing, comparing... That's why it's so slow. That's also why you should try and get some art training, even if it's just a weekly figure drawing class (figure drawing is recommended for any artist, no matter how good).

About Human References

The average eyeball is 24 mm in diameter, with very small individual variations. The average female skull is 21.6 cm high, 19 cm deep and 15.2 cm wide, with a horizontal cross-section across the forehead that is egg shaped (narrower end forward, slightly squared off). The average distance between the pupils is 65 mm for women. These skull measurements of course have fairly large individual variations, except for the eyeball diameter, which is surprisingly constant. The total height of the "idealized" average female body is roughly 7 1/2 heads; the realistic average is more like 6 - 7 heads. The upper teeth are lower in front, and getting slightly higher and smaller towards the back. The lower teeth are set back behind the upper, the lower front teeth much smaller (almost half as wide) than the upper frontals - the result is that the lower side teeth are offset half a tooth's width vis-à-vis the uppers, like two adjacent rows of bricks.

Use drawings, artist anatomy books, illustrator's reference books like "Fairburn Heads" etc, underwear and swimwear catalogs, fashion magazines, health magazines, men's magazines, websites about models or movie stars, your friends... see my link section.

Look for straight poses, shot horizontally from the front and the side, or as close to it as possible. Select expressionless poses in the beginning, until it becomes time for you to start animating the smile and other expressions.

When it starts to look finished, use the technique I call photo matching (can be used for any modeling project) - select a photo, and then match it: focal length, distance and angle to subject, light sources strength and angle, cropping, shading of objects etc. Then start test rendering, scrutinizing and editing the model until the test render starts to look like the photo - it's up to you of course how close you want to try to get. This is a very time-consuming process, since you are constantly discovering new "discrepancies", especially after any sort of break or pause. When you're done with one photo, select another shot of the same object from a different angle, maybe with different lighting, and do the same thing again. And again, as many times as you like, or have reference for.

In the final stages it helps to apply blur and grain and similar background - the whole point of this technique is to make it easier for your eyes to Spot The Difference in the actual model from the original, by isolating it. A very good help is if some of the photos have any highlights and shadows - trying to match highlights is often more accurate than following plan-drawings (IF you can get your lights in roughly the same position as the ones in the photo, which usually is possible). A great book for photo matching faces: "Making Faces" by makeup genius Kevyn Aucoin.

Things to look out for - hardest in the face is the proper curvature (when seen from above or below) of the lips, teeth, forehead, cheekbones/cheeks. The jaw's rotational pivot point should be behind the neck, since the real jaw both rotates AND slides down a bit as it opens. Asymmetry is important in the face (not too much though if you're trying to make a beautiful face, just an almost subliminal amount will do, otherwise the face becomes less attractive to most people), tweak the features at the end like raising one eye, skewing the nose a bit etc.

Hair can be realistically done using complex realistic textures and transparency maps on very simple geometry. To get those maps, you can use other photos, like I did here (I cropped out the areas marked in red, and retouched them). Then I painted transparency maps that crop out the un-wanted parts, by painting on a layer in Photoshop.

 

error-file:tidyout.log

You can also photograph people yourself, or put a wig in a scanner. I don't recommend "Paint Effects" for long free hair - it's great for short hair, and things like tentacles and dreadlocks though.

Cheating Is Necessary

By 'cheating' I mean anything you do in 3d or 2d to create an effect that actually isn't there in your scene. You should be doing it constantly; any time the effect you want isn't possible in your software, in any software, or just very hard or time-consuming. The placement of your objects in 3d space may have to be completely unrealistic to make an image work - never mind accuracy, what's important is - Does it look good? Does it work visually? If it doesn't, you're just wasting your time.

The only thing you mustn't do when cheating, is to get caught.

Real world photographers and cinematographers do in-camera cheats such as using dulling spray, distorting perspective, creating fake reflections, filters, altering their subject, or cheating its placement, as well as retouching and composites on the prints. In film we all know the cars that explode at the slightest impact, the night scenes that are either shot 'night-for-day' or with a huge glaring light hidden behind some trees, the audio that's been added later, the too-perfect actors speaking too-slick lines, sounds in outer space etc... But there's much more that we never think about, because it has become part of the basic vocabulary of film. The whole medium is one huge distortion of reality - reality doesn't direct our attention to things, or have cuts, it doesn't have "viewpoints", or a soundtrack, or a plot. So don't get hung up about that kind of reality - when I talk about 'Reality' in Part 1 I'm talking about capturing the richness and beauty that surrounds us in every direction, in both macro- and microcosm, not about scientific accuracy and objectivity.

The only thing that matters here is the end result, that is the image or the animation, and the emotions it evokes in the intended audience, how they perceive it.

And on that subject, some notes on how we perceive things:

Visual Perception

Why do we perceive some things as "beautiful", and other things as "ugly"? Everybody agrees on which is which, when we discuss certain kinds of things, like sunsets, flowers, horses, jewels, precious metals, a pretty face, babies, clean hair, healthy skin, etc etc... Why those particular things? And does everybody really agree?

Babies have been shown to spend more time looking at attractive faces than unattractive ones, isolated tribes have shown similar reactions. Also more complex figures can attract newborn children's attention for a longer time than simpler ones.

There is an important concept in understanding how our eyes process visual information: the saccade.

A saccade is a special kind of eye movement, usually 'automatic' or involuntary, that rotates the eyeball a very short distance in a very short time - these saccades are very noticeable when you look at someone reading a newspaper, for instance.

Saccadic automation seems to be extremely important to our seeing - in fact without it we are blind.

The area of highest resolution on the retina, the fovea, is of a very small size, about 0.4 mm in diameter. Acuity of the retina quickly drops away from the fovea.

Saccadic automation increases our visual coverage a lot, letting our eyes rapidly scan an area about 10 times the size of what the stationary fovea can see.

In 1950s researchers showed that if the saccades are eliminated, and an image is 'frozen' with respect to the retina, the test subject stops seeing this image in 1-3 sec, basically goes blind. This is called a blank field.

Miner's nystagmus was first mentioned back in last century, in England. It was caused by the miners' permanent visible field being too uniform. A pilot described his vision sensations on flights over the Antarctic: "Imagine you are sitting in a room and for hours are looking at a well whitewashed ceiling". It is quite obvious that such an environment, even in the short term, rapidly causes "visual starvation", and in the long term may cause impairment of the nervous system.

An example of an "agressive field": A wall with thousands of spots on it - the distance between spots about 15 mm. From a distance of 4 meters a visitor sees the width of the picture at an angle of 21.5° , and a space between the spots amounts to 0.23°, roughly the distance of an average saccade. Standing there and looking at it quickly gets extremely uncomfortable. 1960's Op-Art was based on this uncomfortable sensation. If taken too far this can cause nausea, or seizures.

So we want to look at diversity and change and richness. Or to put it another way:

We like variation. In fact we physically need it. And we like varying variation even more.

And we really like varying variations of variations...

And so on...

And we also like details, and fine lines.

Details help the eye stay focused.

Edges are apparently very important to our sight - our seeing is edge-detection-based, on the most basic level (how the photo-sensitive cells in the retina are connected to each other). And lines are nothing but edges... presumably why we like to create and look at line-art so much.

I think of visuals in terms of Levels of Variations, similar to harmonics in music - Level 1 would be totally without any variation, a pure sine-wave; Level 2 would be adding a slight modulation to Level 1, and so on..

Each higher Level of Variation represents a minimal amount of change or modulation to the preceding lower level. (Of course there are a huge number of possible different modulations that could be applied to each level.)

There's alsowhat I call Discontinuities - an abrupt change along a more or less sharp edge.

I think images can be classified according to how many Levels of Variation and Discontinuities they contain.

 

error-file:tidyout.log

Level 0 with no Disc. would be an image consisting of all pixels the same color. Extremely boring to look at.

 

error-file:tidyout.log

Level 1 - there's a noticeable difference on some pixels; a variation, that's repeated throughout the image. A bit better - still boring though. Also, certain variations on this would not be too far from the 'agressive field' mentioned earlier.

.

error-file:tidyout.log

Level 2 is a variation on the variation: in this case maybe adding similar smaller squares, still covering the whole image. Note that here is also added a slight element of randomness, and it's already looking much better than before.

 

error-file:tidyout.log

A Discontinuity would be introducing another area in the image with a different type of variation. In the example above the added Discontinuity doesn't make the image look better, but worse - that's because it introduces a large area of high-contrast Level 1 next to our existing more subdued Level 2, forming an agressive field that almost drowns it out.

We humans seem to like large views - why? This landscape has 4 large Discontinuities (and many smaller ones) of very varied size, color and shape, most of which I estimate are Level 3 or 4, and the borders between them are very varied. It also has a high degree of randomness all over. A higher number of Levels and Discontinuities, varying borders, and high randomness = a more interesting image. I've constructed one with opposite qualities from the first one. Look at them both for a minute or so. Which one of these could hold your attention longer?

 

error-file:tidyout.log error-file:tidyout.log

About Lighting

Again, when it comes to lighting, I think we should strive for realism. If you leave shadows switched off and don't even try them, you might be missing something visually exciting. Or worse, the image can become ambiguous and confusing or just plain flat. Or if you use too many lights and throw too many sharp shadows all over the place... Another 'un-realistic' mistake is leaving the ambient light at its default value, which is way too high. No such thing as the "Ambient Light" exists in real life - a light source that reaches everywhere, or (if 'Ambient Shade' is cranked up) nearly everywhere. Still it's useful to control the darkest darks, so I usually don't just delete it right away, but leave it at a low level, often with 'Ambient Shade' either at default or higher. (Ambient Shade is useful to fake radiosity-like effects, or the effect of very large relatively dim light sources such as the sky, but again if it's too strong it will look bad.) Sometimes 2 Ambient lights each at opposite sides of the scene, both with maximum Ambient Shade, are useful to control darkness and color of shadows.

To achieve realism in our lighting, and also to learn a few useful terms, let's first look at how it's done in the real world:

The basic lighting scheme for film, video and photography is a three-point system, consisting of a key light, a fill light, and a back light. The key light is the primary light in the scene (in an outdoor daytime scene it would probably be the sun). Indoors it's conventionally placed between 30 and 45 degrees from the camera-subject axis and is elevated by 30-45 degrees. Diffuse frontal overhead lighting is the most popular style because it's a fair reconstruction of the general run of natural light and it's efficient. This gives our eyes maximum information about the scene, and is also the most flattering light for people, best showing character and expression. Other more extreme positions might give a dramatic or dreamlike effect, but don't use them indiscriminately. Keeping the key light close to the camera creates a symmetrical lighting scheme, which is also quite flattering for portraits (especially for women, together with a high key setup - see below), but which risks looking boring in other cases. Keeping the key light low creates an unusual and unexpected, confusing, maybe frightening, effect. Putting the key light more or less opposite the camera creates a dark, dramatic, mysterious, intimate, "film noir" atmosphere.

 

error-file:tidyout.log

The fill light is supposed to partially fill in, or soften, the shadows created by the key light. It is a lower intensity and usually more diffuse light than the key light. It is usually placed on the other side of the camera from the key light, at an angle of 30-45 degrees from the camera-subject axis, and at about the height of the camera. If the fill light is too intense, then a low contrast, flat (high key) image may be created. The fill looks good with some color in it, any color if the key light is white, and a contrasting color if the key is colored - but don't overdo it or it'll look cartoony or kitschy, or contrived. Also check - is there something in the scene that might be throwing colored light into that area, like a strongly colored object with lots of light bouncing off it? In that case, use that color in the fill light - unless you judge it degrades the image somehow (see note above on 'cheating'), like if it creates a sickly green effect on somebody's face. If that happens, you might consider changing the color, or the intensity, or the position etc of that object.

The back light is placed above and to the rear of the subject, so that the light does not come directly into the camera lens (not a problem in cg of course). It serves to outline the subject, especially the upper portion, and to separate it from the background. This light also looks good with some color, maybe a warm gold or a pale blue; try to keep it different from the background. The red B in the diagram indicates how the back light usually can and must be adjusted in a cg scene, as opposed to real life.

For large complex scenes, sometimes the background needs extra light, as well as details such as the eyes, certain props, increasing or decreasing focus and contrast in some parts of the image. But avoid having multiple shadows cast from the same object if you can.

As I mentioned before, cheat - you should always strive for realistic LOOKING lighting, but don't worry about physically based lighting too much, it'll get in the way of the atmosphere you're trying to create. The Rembrandt image "The Holy Family" (shown above) is a good example - the lighting looks very realistic, and also extremely beautiful, but if you study it, you see it's physically a bit off. The bright spot on the floor, for instance, doesn't seem like it could come from the same source as that lighting up the main figures, although at first that looks correct. The reason he did that, of course, is to create a more pleasing pattern of light and shade. Remove the bright spot by placing your finger on it, and see how important it is to the composition.

There are high-key, intermediate, and low-key lighting schemesThe high-key image has its curve pushed to the high end, and the low-key to the other. But that doesn't mean that you can change an intermediate image to high or low key in an image editor, by simply editing that curve (changing the gamma for instance).

Conversely, to light your scene high-key doesn't mean to use the same lights and just pump up the power - you have to add more lights, as well, and carefully place and balance them. You're looking for a light, airy effect without any strong shadows or lots of burned out highlights.

Light could be used to select the parts of the image that are more important to you, which you want the viewer to focus on. For instance, if your image can be divided in fore, middle and background (or even more planes), a similar light on all of them will seem a bit flat. So choose one of them to receive more light than the others. Again, Rembrandt and many other classical painters are good examples.

Here's a list of what I usually do to the lighting when I start up a new file:
1. Create an Ambient Light, take the Intensity down from 1 to roughly 0.2 (or change its white color to a dark gray, same thing, except it's more intuitive - the dark gray will then be a visual feedback of what the darkest shadow in your scene will look like).
2. Move the Ambient away from the center to make better use of the Ambient Shade function (where to depends on the other lighting in the scene, maybe to one side or the other, or closer to the camera, higher, lower - I usually start by placing it where I expect some light to be bouncing from...)
3. Place a spotlight in the scene with Intensity a bit higher than 1 (If the spot is the only one, or if its the main spot and no other spots are overlapping it much, it should be intensified to about 1,3, more if you have a lot of Dropoff), switch its depth-map shadows on (never raytrace unless you have to, leave the default Dmap shadow resolution at first for test rendering but maybe raise it later), give it a bit of penumbra, size it so it displays nicely in your scene: size the cone so it just lights up what its supposed to and not too much more, group it to itself and center the pivot in whatever the spot is aimed at (this enables you to scale it to move it back and forth at any angle conveniently, and to rotate it around its target easily).
4. Sometimes a second and third spot will be used right from the start, though it's better tosave the really advanced lighting for later.

The spotlight is the best and most versatile cg light, because it's the most complex. Almost always use it instead of a directional or point light (Point lights with depth map shadows switched on will take much longer to render than a spot with the same shadow resolution). An ambient light in each scene is normally a good idea, though sometimes it isn't. Play with the "ambient shade" setting and placement of the light for more edge-definition - if the light is close to the camera and the "ambient shade" is high, you will see darker edges on all rounded objects.

After placing the lights in a good basic setup, and setting a first ballpark intensity, you can continue setting up the objects in your scene, and adjusting the shading. When you feel the scene is almost finished, you can start thinking about the lighting again. Check out different setups. When you've settled for one (although you may want to save other setups to disk) you can start the fine-tuning: adjust diffuseness, color, whether or not a particular spot casts a shadow, or what objects it lights up, the color of its shadow, fine-tune the spread, penumbra, and dropoff if they are visible in the scene.

Adjusting the diffuseness can be done in several ways. In raycasting, one way is to lower the resolution of the shadows, another is to raise the quality of the shadows - obviously the first is preferable for speed reasons as long as you don't start seeing low-shadow-res artifacts on anything (staircase edges on shadows, and if the Blend Offset is too low as well, blurry spots arranged in a grid on lit surfaces). When you start to see them, raise the resolution, and the Blend Offset. Another way to diffuse the light is to use the spot cluster method - see below. Of course, in raytracing, you can switch on Soft Shadows in the spotlight, or use an area light, but be warned - rendering speeds can slow a lot if you're not careful.

You can project a black and white ramp in the spots color channel to create an improved custom penumbra - the default penumbra has a linear falloff, and setting your ramp interpolation to "smooth" or "bump out" will noticeably improve it. You can also project other custom images and/or ramps to create textured spots, like the uneven circles of light you get from a flashlight, or light filtering through blinds or tree branches, or the clichéd church window...

 

error-file:tidyout.logerror-file:tidyout.logerror-file:tidyout.logerror-file:tidyout.logerror-file:tidyout.log

The first image shows the default look of a spot (Spread 50). No. 2 has Penumbra set to 6. No. 3 also has Dropoff at 5. No. 4 goes back to default values but with a Ramp mapped in the colorchannel ("Bump Out", black at position 0.71, white at about 0.5). No.5 adds some noise and extra banding in the ramp, but this is also by far the slowest to render.

Use negative lights (strength below zero) to remove light in overexposed places, or deepen certain shadows (although the "shadow-surface" method - see below - is faster for rendering).

For more realistic lighting (in Alias) use clusters of spots instead of one - 3, 4, 5 or 9 next to each other, aiming towards the same point (place 1 spot, move its pivot point to the center of the main object to receive the shadow, make x duplicates, rotate 10 - 20 degrees), with their strengths lowered (divide original strength by number of lights) and shadows blurred enough to merge them (usually shadow resolution about 50 - 150, make all of them a slightly different resolution to smooth the 'jaggies' out). Beware of flickering shadows in animation, and shadow artifacts, keep their resolutions not too low, and raise the Edge quality. This mimics an arealight, but it's usually faster to render.

Note - since this was written Emanuel Campin has invented a much better way to get the same effect, although it's limited in that it only works when the distance between light and object remains the same, and there's only one main object. Find his work at Highend3D.com, in the Maya script section.

 

error-file:tidyout.log

This is a cluster of nine, all aiming at the same point, the nearer left bottom corner of the cube. The cluster has been twisted about 45 degrees. The spots are spaced at roughly 20 degrees from each other, and positioned just outside the upper right-hand corner of the image. The Shadow Resolutions range from 90 to 130; the spreads are 45 degrees, the Edge Quality 1, Decay 0, Penumbra 2, Intensity 1,2.

Faking Radiosity and Arealights

Raytracing and raycasting only include primary light paths in their algorithms, not secondary, and uses a constant ambient light as a simplified substitute. (For instance, sunlight through a window is the only lightsource in a room - the sunlit patch on the floor will in turn act as an area light and illuminate the rest of the room.) Radiosity calculates these secondary bounces, and the results are quite realistic, rich and beautiful. But it is very slow if you want high quality, even if it only needs to be calculated once for the first frame in an animation. And faking radiosity using area lights will be even slower. (An area light is not emitting light from a single point like the spotlight and pointlight do, but from a rectangle or circle (that can be scaled). This creates shadows that become blurrier further away from the shadow-casting object, which is often much more realistic than what a single spotlight can do. Alias does this, but it's very slow to render.) So here's how to make your renderings look a bit more radiosity-like: paint more realistic shadows in the color channel of your shaders, or apply extra "shadow-surfaces" on top, and maybe place extra lights (not arealights!) in the scene - basically, PAINT with light. In reality usually at least one part of any surface is lighter or darker than the rest, and if you stop and think for a while (from where is the secondary light coming?) you should be able to figure out which, and how much. Then you either light it, paint it, or use a linear or circular ramp, or a combination of these, either straight on the surface itself, or for more flexibility, apply it as a transparency map in a black Lambert shader on a second surface. I call this a shadow-surface. Here's an example of a painted one that can be used under any table, chair, bed or cupboard etc with four legs:

 

error-file:tidyout.log

Advantages to the last method: you can move it easily and interactively (you can see it in the modeling window) at any time, you can reuse the lower surface's shader on other geometry, and you can reuse the shadow-surface's shader on other shadow-surfaces. Remember to switch shadows OFF for the shadow-surface, and, if applicable, for the object getting its shadow faked. For interior walls, floor and ceiling, a faint circular ramp (darker at the periphery) layered into the color channel almost always looks great. If you judge one side of the room to be much lighter than the other, another straight ramp can be layered on the circular one, and so on.

Here's an example of a good way to imitate sunlight. Use one huge spot very far away, with very high shadow resolution. If a large vista is visible in the image, you can't use that spot for it all - instead, render the distant horizon separately with a no-shadows light, or paint it, or use a photo, and put it in the background of your final image.

The ambient light can be used as the scattered blue light from the sky - place it high, in an opposing part of the sky from the main spot, with Ambient Shade 0.5 and intensity close to the default, and make it slightly bluish. The main spot then has to be made slightly warmer in a corresponding degree, to compensate and keep the sunlit surfaces neutral.

More realistic alternative: use the ambient light for light bounced back from the ground (no blue, a bit darker, placement below the ground), and use either:
1. One or two large spots placed away from the main one, and angled in to shine on the same area, with very low shadow resolution, very low intensity and quite bluish, or
2. one large arealight, low intensity, bluish, placed above so that it seems to come from where the brightest patch of sky visible from the target, where it's going to shine. (Not recommended for animations.)

For the best quality, use Emanuel Campin's method of simulating global illumination in Maya.

Faking Caustics

Caustics are the bright spots and streaks created by the focusing of refracted and reflected lightrays, like for instance the orange highlight on a table underneath a glass of brandy, or the wave pattern on the bottom of a pool on a sunny day, or the highlight inside a crystal ball, or the bright spots on the street created by chrome hubcaps reflecting sunlight etc. (Any curved shiny or transparent object with a strong directional light on it.)

Maya can do this now (download the plugin from the Alias site) but it's a bit slow to render, and tricky to tweak. It can easily be faked by an appropriately colored spot placed inside the object. Look at the image below right - the main light is coming in from the upper right corner. The caustics for the brandy glass can be approximated by an orange spotlight, with position inside the glass and direction like the key light. It should be grouped to the glass for animation, and should always point in the same direction as the key light (this can be fixed dynamically with an expression). The spot should have Decay set to 2 or 3 so it doesn't shine too far, shadows Off for speed, and maybe even an image file mapped in its color channel, for more realism. For a glass of brandy, perhaps something like this (turn it so the circular part points away from the object):

 

error-file:tidyout.log error-file:tidyout.log

When it's mapped in a spot, "overdrive" its color and brightness by setting its Rgbmult V to 3, with a bit of red added in the H and S. The second picture shows the shadow-surface shown above and the caustic image in use, as described. The scene is raycast (Dmap rendered), and renders very fast.

Now that you've gotten free know-how on this topic, try to grow your skills even faster with online video training. Then finally, put these skills to the test and make a name for yourself by offering these skills to others by becoming a freelancer. There are literally 2000+ new projects that are posted every single freakin' day, no lie!


Previous Article

Next Article


Flora's Comment
It's imperative that more peolpe make this exact point.
10 Thu Jan 2013
Admin's Reply: