Add realistic shading
Zeno Pelgrims shares tips and tricks for creating a believable yet stylised image
Create cartoon-like images with believeable details and defects, using Maya, Zbrush and Arnold tools
W hen I saw Andy Estrada’s line work drawing, I got insanely inspired to make a 3D interpretation of it. It lit a spark in my mind. This happens to me when I get my own great idea or when I see someone else’s work.
The image reminded me of my own childhood – being completely engulfed by fascinating things, without any worries. You don’t realise it at the time, but these are some of the purest moments of your life. The image also contains a message to all parents: remember that when you come home from a night out and your little boy is in an absolute state, he might have had a life-defining moment.
In this tutorial I will introduce you to my thoughts and techniques while creating this realistically shaded cartoon-like image. I will skip over the general aspects and instead talk about things I look out for that are especially interesting or useful. I will cover useful topics from the value of cheating to the best look development practices, my post-production tricks to better simulate lenses, Marvelous Designer workflows, how I light my scenes and a whole lot about special shading cases. There’s lots to go over and plenty of helpful tricks for you to learn.
01 CHEAT IF YOU CAN
When you already know a fair amount about the 3D process, it’s not a sin to take shortcuts. Sometimes finishing an image is more valuable than doing it all 100% production-correct. On this project I don’t model anything that’s out of view. So I don’t make the character any legs, and only give texture to visible elements. When you’re just starting out with 3D, though, I specifically recommend that you don’t take any shortcuts, since being able to do things properly as they’re done in production will help you to get hired.
02 RAMP THE EYES
The eyes are the most important part of the whole image, since humans instinctively search a picture for these first. We connect with them emotionally to help us understand a character. Therefore, it’s vital that the eyes contain depth. In real life, depth is created by light hitting the cornea and casting shadows in places and highlights in others. In 3D, especially cartoonlike work, ramps (gradients) are often used to exaggerate the effect of highlights and shadows. I layer these black and white ramps over my original texture in the shader for easy control on a shot-by-shot basis.
03 CRUSH NCLOTH OBJECTS
Next, I create ncloth objects in Maya. I place a metal object such as a can on a floor plane and make the floor a passive collider and the can an active collider. Then I create a solid object to crush the can with, such as a bevelled cube, and make this a passive collider. I select the thin-sheet metal preset in then cloth attributes for both objects. I animate the cube so that it ends up being pushed into the can. I experiment by running the simulation and trying out different initial states to get different end results. I can move the cube or place the base object in a different position, for example, until happy.
04 CREATE POPCORN USING ARNOLD STAND-INS
For the popcorn, I sculpt four variations and scatter them on a proxy surface that I later remove. Because there are so many pieces of popcorn and each one needs to hold a lot of detail, I export the four variations individually as Arnold Stand-ins. To scatter the pieces of popcorn around, I make a low-res mesh and use sppaint3d for Maya, a simple raycast-based scattering script. Once happy with the placement, I select Modify > Replace Objects to swap the low-res meshes with the Arnold Stand-ins. By doing this, Arnold can raytrace millions of them while just loading the four variations onto memory once.
05 FACIAL TOPOLOGY
You can never get too much practice with facial topology. It’s an intricate puzzle, balancing the ability to hold a shape with some deformability related to the expression range. The mouth, jaw and eye areas are by far the most important, since those will move in the most complex ways. This is why I start my puzzle by perfecting those parts, and then connect it all together. It’s super important to start with very few polygons when retopologising, laying out the base loops. Adding extra loops is easy, but managing them isn’t. Starting rough ends in the cleanest result.
06 USE HEIGHTMAPS FOR HELMET DISPLACEMENT
Since the displacement on the helmet needs to look as crisp as possible I decide not to sculpt it, but instead use some mechanical-looking heightmaps and apply them on the shader level at rendertime. Applying a Displacement Map in Zbrush and then exporting both the mesh and displacement again is always destructive to the height data to some extent. So by combining multiple height maps using the allayerfloat, I can precisely position them, while still maintaining all the resolution from the original heightmaps.
07 DAMAGE THE HELMET
Making the helmet look realistic with shading is one of the most interesting aspects of this image. It consists of a metal base layer with a paint layer over the top, which then has a sticker layer above it. I like approaching shaders in a layered manner as it gives me an incredible amount of straightforward control. It would be faster to render if I adjusted all the various maps and used a single surface shader, but often I find that the extra rendertime is well worth it, saving on the set-up time, especially for single frame projects. Key to the layering is the masking in between the layers. Between the metal and paint, I layer various masks such as a curvature shader to add edge damage and a fractal shader to convey overall wear and tear.
08 SHADE HELMET GLASS
On a technical shading level, this is the most interesting shader to create. I use a technique I learned from Lorne Kwechansky. We can create scratches in various ways, such as by using a scratch bump map or a scratch roughness map. However, in this image I use a map to drive the anisotropy instead. This can achieve the nice rings I want, just like with a bump map, but uses a much lower sampling rate. It can often be tricky for a renderer to sample such high frequency details.
09 SPLIT LOOK DEVELOPMENT FROM LIGHTING
Working on the look development of assets and on the creative scene lighting together never works out due to there being two variables – the materials and the lighting. It’s better instead to do the look development of all of my image’s assets in a lighting condition that I’m used to seeing every day, such as in bright sunlight. This is because if you do the look development in a self-set lighting environment, you might end up compensating for the lack of light by making all diffuse components brighter, for example. This will make your image look unrealistic. I develop the look and when I know the materials are spot on, I push the lighting without worrying about the materials any more. Try to eliminate variables in the process!
10 DESIGN FABRIC ELEMENTS
I simulate all of the fabrics using Marvelous Designer, from the t-shirt to the pillows. One of the trickiest things is to get usable, clean geometry out of it. I approach this by exporting both 3D meshes of the simulated 3D shape and the UV shape from Marvelous Designer. This enables me to perform superclean retopology on a 2D plane. I then use the Wrap Deformer on the UV shape and blend shape the fabric back into place on the object. This is by far the most effective way of getting animation-ready topology out of Marvelous Designer that I have come up with so far.
11 CREATE USABLE TOPOLOGY
I export both world space and UV space meshes from Marvelous Designer, as I need both to easily create new topology. First I export the regular 3D object. Then I select the meshes and select Reset 2D Arrangement in the right-click pop-up menu. This transforms the UV space coordinates into 3D coordinates. Then I export this mesh. In Maya, I import both objects and do the retopology on the flattened clothing. I start with extremely simple geometry such as a plane with no subdivisions, and start snapping the vertices to the corners of the flattened mesh. Then I just subdivide and improve. When this is done, I wrap the retopologised mesh to the flattened object. Since the only difference between the two exported meshes
is the coordinates, I can blend shape between the two to put the retopologised mesh back in place.
12 ADD HAIRS AND DUST
One of the final touches is to add dust and fallen hairs to make the image look believable. In any real picture, we won’t notice these until we actively search for them for a while. This is exactly what we want to achieve in 3D too – they need to be there, but they definitely can’t be obvious. For the hairs, I create a couple of curves and place those around the scene. Then I create an Arnold Curve Collector, and render them using a hair shader. For the dust, it depends on how close we are. If we aren’t close up, I use a combination of the Maya Snow shader and an Ambient Occlusion effect to show dust on top of objects and accumulated in crevices. If the dust is closer to us, it’s sometimes important to see the outlines of it. In these cases, I revert back to the curve rendering approach, adding a Lambert shader.
13 SET LIGHTING VALUES
While lighting an image, it’s crucial to get the luminance values right. Luminance is the perceived
brightness, rather than the absolute one. Our eyes are much less sensitive to blue light than green light, and one of the big aspects of creating good lighting is creating depth in the image. This is often done by layering values, so I make sure that layering is obvious. If I can zoom out to a thumbnail and still read the image properly, I know I’m on the right track. To check my values in Nuke, I just press [Y] to enter luminance mode.
14 ADD EXTRA REFLECTIONS
Since the scene is so dark, I have a lot of room to play with nice reflections. I love placing these small-influence lights, since I can really shape the whole image using them. I can dictate which edges need to be highlighted and separated from the background, and which others should blend in. The main ideas that constantly play in my mind when doing this are: “How can I bring as much depth to the image as possible?” and “Which shapes do I want to be read, in order to get a pleasingly composed image?” In general, I’m never afraid to push these things for effect. It often might not feel completely ‘correct’ to place these lights, but when you see people light a real cinema shot, I’m sure you’ll feel much better once again. These ‘cheats’ to make prettier images are used everywhere.
15 INCLUDE LENS DISTORTION
There are a few things we can do to make the image look more natural without the viewer being aware of it. One of those things is adding lens distortions, to mimic the spherically symmetrical curvature of the lens elements. We look at real photos every day of our lives and have become used to the fact that no lines are ever straight. So when they’re straight in CG, we tend to notice it without being able to pinpoint what exactly feels strange. Using Arnold, I use the Fisheye Camera node to introduce a slight amount of distortion to the image. It’s better to do this in-render than in post-production, since no pixel resampling needs to happen in this case.
16 USE EXTRA LENS EFFECTS
After the image has been rendered and I have done all of the AOV compositing adjustments, it is time to add in the lens effects that are not easily simulated by a raytracer. For example, due to the curved lens geometry and the wavelength-dependent index of refraction of most materials, real images have several lens aberrations, of which chromatic aberration is just one. It’s important to add chromatic aberration carefully, as it’s a depth-varying effect. Colours are switched before and after the focal plane, and the effect gets more intense the further away the pixel is from the focal plane. It is also common to find chromatic aberration around the edges of an image, even when those points seem to lie on the same focal plane. However, again due to the curvature of the lens, the focus point will actually be at a different distance over there.