An introduction to game model skinning

copyright © 2004 karl g nyman




Skinning prerequisites:
NOTE: Professional 3D-painting programs like Deep Paint 3D and associated tools like Deep UV are designed for production-level skinning tasks, but are priced out of the general hobbyist's range. This tute presents techniques that are more labor intensive but can be accomplished with software that is low-cost or free, or likely already in a beginning skinner's toolbox.

Since this tutorial is intended to be non-application-specific, there are no tutorial prerequisites -- just reading the tute will benefit many beginning skinners. But to actually perform the skinning operations, there are a few "must-haves":

  • A 3D modeling app that can generate a UV texture map (e.g., 3DS Max or Maya). The freeware versions, gmax and Maya PLE, can be used. For skinners working on models made by other artists, an "unwrapping" app (e.g., Lithunwrap or Ultimate Unwrap3D) is essential if an appropriate modeling app can't be used.
  • A sophisticated image editing program (e.g., Photoshop or PaintShop Pro) with layering capabilities is highly recommended -- rudimentary painting progs (like Windows Paint) can be used, but complicate the process a great deal.
  • Drawing skills of the pencil & paper sort are extremely useful, the more advanced the better, both for planning a skin and preliminary execution of the skin painting. Digital drawing tools can be used instead, but may require advanced proficiency. Hand-drawn art will require a scanner for digitizing.




Overview

This tutorial is designed to provide information regarding the skinning process. While the majority of information in the tute is useful regardless of the programs used, some aspects of skinning must needs be application-specific, and the exact steps for every possible program that can be used will therefore be different than shown in this tute. However, the steps in the process do illustrate general principles that can be extended to specific cases with a little effort.

FIG. 1 A human female model and its skins, used in part of this tutorial.




Some basic explanations of terminology

"Skinning" a 3D model is the process by which the visible surface features of the model are planned, painted, and affixed to the model mesh. Sounds pretty simple, right? Not even. Skinning can be one of the most labor-intensive tasks in the game-model production process, because it usually involves taking a complex 3-dimensional surface and "flattening" it out in such a way that it can be "mapped" to a 2D image, or, as it is more often called, a texture.

Calling this process "mapping" is not just an analogy -- in practice, the area of a rectangular texture is the plane upon which a special map of the model's surface is drawn, so that every point on the surface has a corresponding position on the texture's image plane, much as the surface of the earth is portrayed by contours on topographical maps. As with terrestrial topo maps, textures (also called texture maps) for game models follow a coordinate system when applied to the models, and an understanding of that system is important for skinners.

U V W...whazzat ???

Textures on models follow a three-dimensional system with mutually perpendicular axes usually labeled U, V and W. Although this system is similar in some respects to the XYZ cartesian frame of the virtual 3D world, there is no actual correspondence between them. If one considers XYZ the "model space," UVW would be model's "surface texture space."

FIG. 2 UVW axes at various points on a model's surface

There's a very big difference between these two spaces: The XYZ frame is rigid, the UVW frame is elastic. This is important, because unlike the XYZ frame, the UVW frame has to be "unfolded" in the skinning process, so it can be projected onto a 2D texture plane.

So what is the correspondence between the 3D UVW system and a 2D texture's pixel(x,y) system? In general, U corresponds to the horizontal pixel(x) dimension, V to the vertical pixel(y) dimension, and W to the surface "normal" -- a vector directed "off" the 3D surface at right angles to U & V at some given point, which describes how the 2D texture is "projected" onto the surface at that point. For this reason, mapping a 2D texture to a 3D surface is sometimes termed "projection mapping" or "texture projection." However, don't think of the texture as simply being cast on the model by a slide projector -- this is analogous to only one of many UVW mapping schemes.

NOTE: Throughout this tutorial, "mapping" and "projection" are generally synonymous, with some exceptions that should be obvious in context.

Planar, cylindrical, spherical, and beyond

The complexity of skinning arises when one considers the topology (shape in 3D space) of a game model, such as a humanoid player -- all the gross and subtle ins, outs, convex- and concavities of a humanoid form must be represented on a flat, rectangular surface! Even as basic a shape as a sphere can present real problems to a skinner -- it has a continuous surface that is finite but unbounded, nearly the exact opposite of a 2D texture plane (discontinuous and definitely bounded).

Obviously there has to be some compromise in the "translation," and this is where the art of skinning requires a somewhat scientific eye. As mentioned, the UVW coordinate system is elastic and plastic -- it can be stretched, squeezed, and otherwise distorted to obtain a best fit on the texture plane. It can even be broken into components and stitched back together to accomodate the texture. But such UVW manipulations can distort the texture once it's re-applied to the model's surface, so choosing the method of fitting the model's UVW system to the texture's pixel(x,y) system (or vice-versa) is where the skinner must call on some familiarity with solid geometry.

PLANAR MAPPING
The most basic method of applying a 2D texture to a 3D shape is planar projection (FIG. 3). This is the "slide projector" technique. UV corresponds directly to pixel(x,y), and W (here it is the same as the projection axis) is parallel at all points on the surface. This is ideal for flat and useful for gently curving surfaces, but can produce extreme distortion of the texture ("texture stretching") on highly curved surfaces or surfaces at a significant angle to the projection plane.

FIG. 3 Planar mapping applied to various surfaces

CYLINDRICAL MAPPING
For surfaces with constant (or nearly so) curvature in only one dimension (U or V), cylindrical mapping is frequently used (FIG. 4) -- this technique "rolls" a flat 2D texture so it better matches the curved contour of the target surface. Cylindrically projected images can wrap entirely around an object, and its opposite boundaries can meet, requiring careful texture creation to insure an invisible "seam" at the juncture.

FIG. 4 Cylindrical mapping on various objects

SPHERICAL MAPPING
Regular curvature in both U & V produces spherical texture projection (FIG. 5). In the extreme, two opposing texture boundaries can meet along a line of "longitude," while the other two converge on themselves at the "poles" of the spherical surface. W is described by the normals at the sphere's surface. Spherical mapping can produce extreme texture distortion at the pinched poles unless specially "pre-distorted" images are used.

FIG. 5 Spherical mapping with a grid texture

OTHER MAPPING
"Box" mapping is a special form of planar projection that uses six projection planes that correspond to the faces of a cube that encloses the model. In theory, the "box" can have any number of sides equal to or greater than 4 (a tetrahedron), but in practice, twelve is a common maximum, and the term "box" is usually reserved for the six-sided case. In Maya, this form of mapping is called "Automatic."

Other specialized mapping methods exist, such as assigning a texture to each face of a model, but are not often used in game model skinning.

So far most references to "projection" and "mapping" have given the impression that the texture image is somehow "stretched" to fit the model surface. Actually the opposite is true -- the plastic/elastic UVW coordinate system of the model's surface is "unfolded" in a regular fashion to get it to fit with as little distortion as possible on the flat plane of the texture. A point on the image's pixel(x,y) system is thus made to correspond to a point on the model's UVW system. It's when the UVW system, along with its "mapped" image pixels, is re-applied to the model surface, that texture stretching can occur. So the primary task of a skinner is to produce a flat UV map from a 3D surface that distorts the metrics of that surface as little as possible.

NOTE: Once a model's UVW coordinates are unfolded & flattened, the W dimension can usually be ignored, hence the term "UV map".




The tools of UV mapping

Proper skinning demands topological manipulations. Topology is a twisty branch of mathematics. Mathematics deals with numbers. Computers are really good at doing numbers. So... let's let COMPUTERS do the UV mapping!

All well and good, and for the most part the best route, except that 'puters need to be told EXACTLY what to do, and each and every model is, if not a duplicate of some other model, unique, so there is no universal formula or algorithm that will generate a perfect UV map every time. Here's where the art of skinning begins to overtake the science.

Most higher-end 3D modeling apps (e.g., 3DS Max and Maya) provide tools for generating UV maps based on the projection methods described above, and some have even more tools for the task. But all usually require finessing by the human eye, mind, and hand to be fully successful. This is usually done in a work enviroment called a "UV Texture Editor" or "Texture coordinate editor," or by some similar label. The common task of these editors is to provide an editable representation of the model's UV coordinates projected on (mapped to) a 2D image plane. The W dimension is suppressed in this workspace, and the image plane can be "loaded" with a file-based texture so the UV map overlays the image to be applied to the model.

In this tutorial, the Maya 4.0.1 PLE (Personal Learning Edition) UV Texture Editor window is used. It has a very clear and understandable GUI and a wide assortment of tools for manipulating UV maps.

Mapping and skinning a model with the basic projections

Relatively simple model shapes produce simple UV maps (FIG. 6). Here the main body of a missile model has been mapped with Maya's default cylindrical projection (which matches the model's essential shape -- it began as the cylinder shown along side the missile), and the flattened UV map is displayed. The image applied to the model is shown in the background of the UV Texture Editor window.

FIG. 6 Default cylindrical mapping on a model reshaped from a cylinder

The grid texture used makes it easy to see the distortions to the default mapping caused by the modeling of the missile -- the vertical spacing of the gridlines on the missile is no longer even, because the space between mesh vertices has been altered. By manipulating the UV map, this distortion can be remedied.

In FIG. 7, two UV maps are shown, that of the original cylinder (on right of the UV Texture Editor window), and the adjusted map of the missile model (left). The vertical spacing between rows of UV coordinates on the model has been adjusted to make the texture match the regularity of the original cylinder. This was done by selecting the intersections of the UV map (which are commonly called UVs) and moving them along the V axis only (i.e., vertically in the orientation shown).

FIG. 7 Moving UVs to adjust texture regularity

Editing UVs is similar to editing vertices on the model's 3D mesh, except the coordinate system is UV (with suppressed W) rather than XYZ. In the more sophisticated editors, UVs can be moved, scaled, rotated, cut, pasted, stitched, and more.

Note that though the missile's adjusted UV map extends beyond the image area in the UV Texture Editor window, the texture is continuous on the model -- it "tiles," with the texture repeating itself in the areas outside the image area. It is possible to confine the UV map to only the image area, but for this regularly repeating texture, tiling is acceptable.

FIG. 8 Texture distortion on cylinder caps

FIG. 8 shows the cylinder & model from a top view. In this case, the end faces (caps) of the cylinder & missile show some undesirable texture distortion -- the cylindrical mapping isn't the best method for these surfaces. Instead, a planar projection along the Y (vertical) axis is applied to the caps, then the resulting UVs are manipulated with scale & move to fit the texture pattern of the cylindrically-mapped faces (FIG. 9).

FIG. 9 Remapping the caps

In the UV Texture Editor window, note the multiple groups of UVs distributed on the image area (missile on left, cylinder on right). This is a common method of using a single texture to map a number of different UV projections on a single object. The overlapping UVs are OK because of the regular grid pattern of the texture. Other textures (such as insignia and other marks applied to the missile) might require that the UV's not overlap.

The missile's guidance fins are mapped separately using planar projection. This will produce a separate UV map for the fins. To combine the fins UV map with that of the rest of the missle, the meshes for the components are combined into single mesh, and the UV map components then selected and moved to separate parts of the image area (FIG. 10). In this way, a single texture can be mapped to many different parts of the model (FIG. 11).

FIG. 10 Missile UVs relocated in image area

FIG. 11 Views of the completed missile texture (left), and with UV map overlayed (right)

This is a critical point for new skinners to absorb -- don't try to always map an entire model with a single projection method. Use that which best suits the model's geometry, and remember that every model can have more than one (and often many) UV maps for its many components, or even for relatively small portions of the same complex, continuous surface. Most modelling programs/UV editors allow some method of combining the component map projections into a single UV texture map, if that is desirable for a particular model.

The creation of the texture file for this model used the UV map from Maya PLE as a template -- once the UV's were distributed within the image area, the UV map was recorded in a screen capture (Maya PLE does not permit UV map exports, although the full Maya does). The captured image was opened in Photoshop, and the texture painted with the usual tools, using the UV map image as a guide to where the various texture elements were placed in the image. FIG. 12 shows the fully skinned missile model.

FIG. 12 The missile, fully skinned

A note about skin "painting": whether the image used as a skin is actually painted (i.e., with an image editor's various tools) or not is irrelevant. Often scanned images, images exported from pre-lit and rendered high-poly models, CG textures, and hand-drawn art & touchups are all used to create skin images. "Painting" is a general term intended to incorporate all these possible techniques.




Mapping and skinning a complex model

Objects like the missile lend themselves to relatively simple UV mapping and painting methods, and are good starting subjects for beginning skinners. Eventually, though, a skinner has to tackle the much more complex task of working with a character model, one of the most complex subjects in the field.

Whether the model is produced by the artist doing the skinning, or by another member of a mod team, or downloaded from a public website -- the source isn't particularly important -- the first step is to thoroughly study the model's contruction, analyzing its components, in order to plan the skinning process ahead of time.

Some questions to be asked and answered in studying the model include: Can and should the model be optimized further? Changing the mesh after skinning is not always a good idea, though most higher-end apps permits this. Is the model a single continuous mesh, or made up of separate components/segments? Will the head and face require separate mapping and painting? Are there model features that require "special treatment" due to complexity, such as hands & fingers, eyes, mouth & teeth, etc. Is the model (or any portion of it) bilaterally symmetrical? Where on the model can the inevitable texture seams be best hidden, or made as invisible as possible?

It's often helpful to sketch the model from a number of views (if this hasn't been done already prior to or during the modeling process), making notes on the sketches about model details that might affect the UV mapping process. These sketches can also be useful when planning the skin painting. The more decisions can be made ahead of time about the final look of the model, the less time will be spent on the actual mapping and painting, which requires mechanical skill as much as creative talent.

For this tutorial, a female character model, Alisan01, will be the example (FIG. 13). This model presents typical skinning considerations: well-modeled facial features, humanoid body structure, and some form of costuming. The model is sufficiently optimized, and has two main parts -- the body and the head; the head will be mapped separately from the body. There are no special-treament features, the body is bilaterally symmetrical, and seam positions are yet to be determined.

FIG. 13 Alisan01, a complex skinning project

Skinning a character for a game (be it humanoid, alien, fantasy creature or what-have-you) is a significantly more complex task than skinning a relatively simple geometric shape such as the missile. This is largely due to the complex topology of the form being skinned, which presents an immediate question: what projection can be used to map the UVs, when the shape is composed of many different 3D shapes with few common planes?

The answer is that no one projection technique will suffice. In general, there are two approaches to UV mapping complex surfaces:

  1. The form can be analyzed and various projections (planar, cylindrical, spherical, etc.) used to create individual UV maps for each segment. This provides a measure of flexibility, but does not keep the component UV maps in scale to one another, since each set of UVs is mapped to a separate image space. Thus a spherical mapping of a head and a cylindrical mapping of an arm would be fit to the same texture image dimensions (such as 1024 x 1024), even though the components are not of similar scale in the model.

    This process also requires either separate texture files for each mapped component, or some method of combining the various UV maps and adjusting their scales to fit into one image area (as was done with the missile components).

  2. The entire model can be mapped at once, using a number of projection planes that surround the model ("multi-planar projection"). This is similar to the box method, but the number of planes from which the projections are made is variable over a typical range of from four to twelve planes. In Maya this is called Automatic Mapping. The terminology in other apps likely varies, but the result is the same -- a single UV map for the entire model is generated in one step. This is beneficial because the scaling of the UVs for the various model components will closely match that of the model by default, and the map will be generated to fit a single image area.

    Typically this process is used to generate two UV maps for each character model -- one which maps the body from the base of the neck down, including all limbs, and one which maps only the head and neck. This allows for greater detail in the features of the head and face than a single map for the entire model would allow. Of course, any number of such projections can be made, based on the demands of a particular model.

The multi-planar projection approach has a major drawback -- the UVs generated from one projection plane are not necessarily connected to those generated from a different plane, though the portion of the surface being mapped may actually be adjacent. This is an unavoidable consequence of this projection method, but also serves to minimize the distortion that is bound to occur when any UV projection is generated.

Keep in mind that the UV mapping process is intended to represent a surface of complex curvature in three dimensions, on a flat surface of only two dimensions -- a task impossible to accomplish without distortion to some degree. Part of the skinner's task is to avoid this distortion as much as possible in the final skinned model. Multi-planar projection is the method most likely to accomplish this end, but it requires a significant amount of manual adjustment to work properly.

This part of the tutorial will focus on the multi-planar method, and uses Maya's Automatic Mapping tool. Rather than describe the exact steps in Maya that produce this type of mapping, this tutorial describes the result, which should provide the information necessary for a reader to adapt the process to his or her particular 3D modeling app.

FIG. 14 The model's head showing polygon structure and fully skinned

The model is being mapped in two parts, head and body. FIG. 14 shows the Alisan01 model's head with a monochromatic material applied and fully skinned.

FIG. 15 Automatic Mapping (Maya) of the model's head

FIG. 15 show the "raw" UV map generated by Automatic Mapping of Alisan01's head, using 6 projection planes. Some portions of the map are immediately recognizable -- facial features, the sides of the head, the top of the head -- but there are also quite a few bits and pieces arranged throughout the map that are less obviously assignable to any one portion of the head.

Looking at the map, note that in many places the UV map structure closely mimics that of the model's polygon faces. In Maya, the UVs are defined as the intersection points of the UV map's polygonal structures (FIG. 16), and in the initial mapping they correspond closely to the vertices of the model. It's important to keep in mind, however, that the UVs can be manipulated independently of the model's polygon structures: moving a UV does not move a corresponding vertex. Moving vertices changes the actual surface of the model. Moving UVs changes how a texture is placed on that surface.

FIG. 16 The model head UVs (green dots)

Why all the pieces? The Automatic (multi-planar) Mapping has taken each polygon face from the model, used the best-fit projection from the six planes specified, and placed that face onto the 2D surface of the image area. In some cases, faces share a common projection plane, and can be left connected -- the frontal facial features are recognizable for this reason. Note, however, that the sides and underside of the nose (selected UVs in FIG. 17) are not part of this set of connected UVs -- these faces have a different projection plane, and are thus placed in the image area as separate sets of UVs. The significant curvature of the face in the nose area has been successfully translated to a 2D surface, but at the price of separating the different UV components of that surface.

FIG. 17 The separated UVs of the nose

This presents a problem: though the flesh of the entire face is a continuous surface, the UV components have been split into separate pieces. The image used to texture such a UV projection would have some parts of the face painted in one area, and others in a different area. Getting a smooth transition between such separate pieces of the surface presents major difficulties.

The solution is to re-position the various UV pieces so they are more closely aligned to the other UVs of the same surface, then combine them into a single UV network. This process is called by various names -- stitching and sewing are commonly used -- but the process is the same. The various pieces of the UV projection are painstakingly reassembled to provide the skinner the optimum "map" upon which to actually paint the head features, and at the same time not introduce unacceptable distortion.

Some new distortion is inevitable. Remember that the initial Automatic Mapping was intended to minimize (but not eliminate) distortion. Since the sewing process must change the initial mapping, some new distortion will occur. Part of the skinner's art is knowing how to minimize the distortion when sewing together an optimized UV map.

FIG. 18

FIG. 19

FIG. 20

FIG. 21

FIG. 18-21 Sewing the nose UVs

Using the nose UVs as an example, the steps for stitching them into the face UVs are shown in FIGs 18-21. First the smaller pieces (sides and underside) are moved into approximate position (FIG. 18). Next, the UV's of both the face and nose pieces are moved to provide a rough fit between the components (FIG. 19). This is where additional distortion is introduced, so the moves must be made carefully to minimize it. Lastly, the pieces are sewn together -- in Maya, this is done by selecting the common edges of the UV structure and using the Sew UVs command in the UV Texture Editor. FIG. 20 shows one side of the nose and facial sections sewn, the other with edges selected, ready to sew. Other apps may have different methodologies or nomenclature, but the result will be the same -- a continuous UV map for the frontal facial features (FIG. 21). After sewing, the UV network is often tweaked to optimize the painting area and minimize texture distortion -- always a matter of compromise between the two.

FIG. 22

FIG. 23

FIG. 22 & 23 Sewing the face sections

As a second example of stiching/sewing UVs, the sides of the head will be attached to the front facial features UVs. In this case, the projections for these parts of the surface are at near right angles to one another -- connecting such projections introduces distortion in the "spliced" area, but by careful pre-alignment of the sections (FIG. 22) this can be minimized. FIG. 23 shows the selected edges sewn together.

In like manner, the back of the head can be sewn to the top. Obviously there are some UVs which cannot be fully stitched together without introducing extreme distortion -- the top and back of Alisan01's head and the rest of the features are an example. Likewise the ears (probably the most complex surface on a head) may not lend themselves to complete stitching. Each model will have its own demands, and it's the skinner's task to decide how much sewing to do. Unsewn UVs will produce texture seams, and the texture painting must take these into account.

FIG. 24 The final face/head UV mapping, fully sewn & ready for painting

FIG. 24 shows the final UV mapping for Alisan01's head. Note that portions of the ears and the top of the head are left separate from the remaining stitched UVs. For the ears, trying to combine them into the main network of UVs would have led to unacceptable distortion. The edges of the ear UV sets, where texture seams will occur, were chosen to be basically invisible on the model, on the inside edge of the outer rims of the ears. Careful painting of these seams also helps further disguise them.

The top/back of the head was left separate for two reasons. While it would be possible to stitch this section to top of the face UVs at the front, that would gain little in terms of surface continuity (the sides couldn't be stitched without excessive distortion), and furthermore would make the proportions of the entire map unsuitable for the image area (1024 x 1024). Instead the top/back section was left separate, rotated 90 degrees, and placed as shown. The seams between sections are disguised in the painting process. This illustrates another consideration for skinners: the UV mapping must fit within pre-defined image area boundaries, and at the same optimize the available image area so the painting can be as detailed as possible.

One preliminary technique used for painting Alisan01's facial features is to print out a version of the UV map in black on white, then place a piece of tracing paper over the printout and draw the features, using the printed UV network to place the features properly. A mechanical pencil with 0.5mm #2 lead was used, and the initial drawing was rather sketchy, as it was only a preliminary layer to be painted over eventually. Register marks were added corresponding to specific UV locations on the UV map. The sketch was then scanned into Photoshop and fit to the UV map using Transform, aligning the register marks for a very close fit.

FIG. 25 Major steps in the head/face skin painting process

Painting the face/head texture was done in Photoshop, using the airbrush, brush and smudge tools primarily. Frequent tests on the model itself were made during the painting to check fit and proportion, and to tweak the details. The animation in FIG. 25 shows the major steps in the progression from an intial sketch to a full skin.

Since game lighting is often inadequate for many reasons, and because poly count limitations make modeling fine details impractical, it's often a good idea to paint in the gross and subtle shading that makes a character model seem much more "realistic." In a face, cheekbone, chin, nose and forehead highlights are paintable, as are the nostrils and fine eye details (lids, lashes, tear ducts, etc.). The interior of the ears, convoluted as they are, demand painting -- modeling them properly would be a severe waste of polygon resources, and if well-painted, can be much more convincing. The goal is trompe l'oueil, or "fool the eye," and all painter's tricks are valid!

For faces, eyes are very important, so pay close attention to their painting. Make sure the "wet" highlights are included. If possible do not simply mirror one side of the face to the other -- purely symmetrical facial features never occur in nature, and often look somewhat unnatural. Once the main painting is done, if the features are mirrored, go back in and create some small asymmetries (e.g., details in the iris and lip texturing, small changes to eye size on one side, etc.). The effect is subtle but much more convincing. In particular, do not simply mirror the bright wet highlights of the eyes -- these are never mirror images of one another (though many commercially-produced skins show them that way -- laziness!).

The techniques used for Alisan01's head and face were also used in the body skinning, but in this case many more individual "pieces" were possible, given the nature of her uniform. FIG. 26 shows the raw Automatic Mapping, and FIG. 27 the stitched final UV map. FIG. 28 is the flat coloration, and in FIG. 29 is the final fully-painted texture map. Notice the use of airbrushed highlights and shadows to enhance the form of the model. This is similar to pre-lighting a model and creating a lightmap, but relies on the artist's skill rather than a rendering program.

FIG. 26

FIG. 27

FIG. 28

FIG. 29

FIG. 26 - 29 Steps in the body skin painting process

The textures for the uniform sections were created as separate Photoshop files, then added to the full-body texture file and masked as needed. Frequent tests were made to identify unacceptably visible seams and to check the overall painting on the model. The uniform textures and details were kept relatively simple in Alisan01 so the effect of transferring the flat painting to the 3D model could be more easily seen. In a "production" model, a lot more detail and seam masking would likely have been added.

You should have noticed that the UV map and the texture for Alisan01's body is for one side only. This is a common technique for models with bilateral symmetry -- model and skin one side of the form, mirror the geometry (and its UV mapping) across the axis of symmetry, then join the two halves together. For asymmetrical detailing (badges, weapon holsters, that sort of thing), some extra polys can be spent on modeling these "attachments" and separate skinning done for them.

In summary, it should again be noted that the techniques described in this tutorial are not the only skinning methods. As mentioned at the outset, there are 3D painting apps that permit painting directly on the 3D models (the UPaint freeware that came with UT2003 is a limited-capability example of this). In the full version of Maya, a 3D painting tool can be used (the tool is also useable in Maya PLE but the painted textures are always watermarked with every save -- not a good thing after hours of painstaking 3D painting!). But for non-commercial game model skinning, where costly investment in software tools isn't common, the methods described work well and are employed as matter of course.

 

chip nyman • january 2004 • chipartist[at]metrocast[dot]net • Home