I’m not much of a modeler and I’m even worse when it comes to art directing my own work. That might turn out to be a good explanation for why it took so long to finally settle on a final model for the Spider character in my animation entitled, simply, “Spider”.
Let’s break it down chronologically, shall we?
I started by finding as much reference material as I could on the physical structure of spiders. Macrophotography, anatomy illustrations, other models and even comic art was gathered. For my own purposes I was specifically interested in the ratio to abdomen to thorax, the spread of the legs around the body, and the structure of the eyes. It was in tis period that I realized that spiders not only have 8 legs, but also have 2 arms or ‘pedipalps’ in front of their head.
There are a LOT of different species of spiders each shaped and colored differently. Throughout my research I came to realize that spiders can be categorized by their method of catching prey (webs, traps, ambush, etc) and by the type of webs they weave (orb, funnel, tangles, sheets, tubular, etc). I wasn’t much interested in trying to model a particular species of spider, but I did have to make some decisions on anatomy, for example the number, size, and placement of the eyes. Most spiders have pretty terrible eyesight, especially the variety that use webs. Only jumping spiders or spiders that hunt by stalking their prey rather than laying a trap for it have good eyesight. Because many jumping spiders have an exceptionally large pair of frontal eyes (perfect for emoting) I chose to style the eye arrangement of my spider off that of a jumping spider.
My spider was meant to be a prototypical spider, not a specific species. If an Arachnologist approached me with complaints that’s one thing, but if a normal audience is distracted by the anatomy or species of the spider, then I’m obviously not doing my job as a story-teller. As my dad always says, “If they’re noticing that, we’ve already lost them.”
Here is a really gross piece of concept art I did before I started modeling:
And here is the slightly less gross model that resulted:
When I began work on Spider, my original intent was for the project to be complete within a few months at most. It was meant to be a layout exercise with little emphasis on the rest of the 3D animation pipeline. This ‘Proxy Spider’ was created as a stand in for the eventual ‘low-poly’ final model. The Proxy Spider has very little in terms of detail, but it adequately conveys the spider’s basic form. The legs were particularly basic serving only to allow for animation tests. At this stage I was still hoping to eventually add fangs and mouth geometry for an eventual scene involving the spider biting a peach. The legs, arms, abdomen, and thorax were all modeled together as a single piece and then later re-modeled as separate pieces floating apart from each other. I still have misgivings about the decision to leave the pieces of the spider floating eerily in space with no connecting geometry, but now, as then, I feel it has a certain appeal that I don’t want to let go of.
The longer I worked with the Proxy Spider, however, the more I realized that it was going to take a lot more work to take what I had and make it look purposefully and artistically ‘low-poly’. The Proxy Spider had a low number of polygons, sure, but it also looked like I had spent perhaps 2 or 3 hours building it. The low poly art that I was seeing on the internet looked purposeful. How could I imbibe my Spider with that same level of intentionality and still finish spider before the semester ended?
The answer was that I couldn’t.
After several meetings with a fellow VIZ Student, Brian Smith, who I’ve talked about in previous posts, I recruited him to re-model the spider. Brian was the closest thing to a low-poly art aficionado I had met thus far and he advocated for a dirty, chaotic modeling style which emphasized purposefully un-clean geometry. Notice how the so-called “Brian Spider” has no recognizable edge-flow or organization to its surface. The polygons have the appearance of shattered glass or, with the proper surface shader, a finely cut gemstone. Brian was interested in low-poly art that rode the line between the beautiful and the grotesque. As he saw it, the spider was a perfect vehicle for his aesthetic, both delicate and intricate, but also repulsive in nature.
Every now and then while presenting the new model to me for review Brian would show me a little of his technique for achieving this style of low-poly art, which is avant-garde even within its own genre. Brian used a combination of reducing, subdividing, smoothing with the sculpt geometry tool, and manipulating vertices and edges. He often started with some previously existing geometry (in this case my Proxy Spider) and then purposely created points of high valence (vertexes with more than 4 edges issuing off of them) by creating extra edges. By reducing and then subdividing the geometry he made maya do a little of the simplification work for him, but in most cases he would simply scour the mesh with his eyes looking for places he felt needed modification, adding or reducing complexity. Though I never quite knew his methods for deciding when and where to make changes, the tools he used and how he used them would eventually help me pick up where he left off.
Here are some earlier screen shots of the model, showing off an highly specular blinn texture that shows off the facets better than a typical lambert. The Eye model was changed in this iteration to become two nested pieces of geometry, both in Brian’s shattered geometry style. The inner eye would move around while the outer eye stayed static, hopefully producing an insectoid eye effect.
For a time this Spider was the final version, even going so far as to have it rigged and incorporated into the animation and layout tests, but my professor and committee member Ergun Akleman had other plans.
Ergun Akleman is a computer graphics researcher with a background in Electronic and Computer Engineering and within the Visualization Sciences program here at Texas A&M, he is an authority on 3D modeling, rendering, and topological mesh modeling. Ergun’s take on the Brian Spider was that the purposefully hap-hazard edge-flow gave the model and distinctly unprofessional look. The model had succeeded in looking intentional in its design, but not in being carefully crafted. In particular, Ergun noted the legs and how the level of complexity drastically decreases as you move from the base to the tip, as if the artist got tired half-way through. “Anyone can make a model that looks dirty. Making a dirty model is easy. It takes hard work to make something that looks clean.” Ergun was pushing me towards the opposite side of the low-poly art fence from Brian, towards the grand-daddy of the low-poly aesthetic, Timothy J Reynolds.
Reynolds’ style seems to emphasize organized chaos (as far as poly flow is concerned), color, and light. Reynolds had called his work ‘3D Illustrations’ and says that the term ‘low-poly’ is a misnomer. His work actually has a lot of polygons, but the style emphasizes them rather than disguises them and in contrast to Brian’s style, works to make the finely cut models beautiful rather than purposely repulsive. In many ways Brian’s style is a rebellion against the Reynolds style, and thus a style shift like this feels like a 180-degree turn. However, I respect and agree with Ergun’s assessment of the Brian version of the spider model and chose to restart for a second time.
Low-Poly art is quietly becoming all the rage and a little bit mainstream (much to Brian’s displeasure) and as a result, there are lots of tutorials online for how to make low-poly art. Some of these tutorials are valuable, and others are not, especially when considering the limitations of the programs I’m working in. For myself, the best way to learn was through observation and experimentation.
From studying Timothy J. Reynolds work I developed a few rules for how the model should look when I was done:
- The model should be made entirely of triangles, no 4-sided polygons or greater
- The model should tend towards vertices with 6 edges branching off
- The edges of the model should be hardened so that the facets of the individual polygon faces are clear
- Complexity should be kept to a minimum, give the impression of an object, not than every little detail
- The sizes of polygons should be kept relatively similar throughout. Don’t put too much detail or too little detail in any one place.
Along with these general guidelines, I also discovered that most of Reynold’s work seemed based off of a single geometric shape, the icosehedron.
An icosahedron is a polyhedron with 20 triangular sides. It is also a platonic solid (regular, convex polyhedron with congruent faces of regular polygons and the same number of faces meeting at each vertex) along with the Tetrahedron, Octahedron, Dodecahedron, and the cube. For all you D&D fans out there, the platonic solids are special in that their congruent (equally sized) faces make for perfect multi-sided dice. The common 20-sided dice used in Dungeons and Dragons games are all Icosahedrons.
Apart from being cool, Icosahedrons (and other platonic solids) easily maintain vertices with consistent valence when you subdivide them in Maya, unlike the sphere primitives that Maya generates for instance.
In the above image two different spheres of different complexity are triangulated and smoothed using the sculpt geometry tool. Notice how the tops of the spheres start as triangles and don’t change when triangulated. Also notice that the top and bottom vertex of each sphere often has higher valence than the rest of the sphere. This can be nigh-impossible to remove. The complexity of the top and bottom vertexes has to go somewhere and redistributing it by manipulating edges and vertices tends to make the whole mesh look worse as a result. The Icosahedron on the other hand doesn’t take much effort to look both organized and chaotic and has a relatively consistent 5-6 valence level. Most of the new model is derived from Icosahedrons or other platonic solids.
The general process was to start with an Icosahedron (or other platonic solid like a cube) and model a section of the spider. If I started with an Icosahedron then the next step was to smooth and manipulate the geometry until I was satisfied with the complexity and the overall geometric form. If I started with a cube I would triangulate, smooth using the sculpt geometry tool, and adjust vertices and edges a bit to give it the same hexagon-like polygonal look that an Icosadedon has.
Above you can see the process done on the legs and arms of the spider.
After much wailing and gnashing of teeth I was able to come up with this new low-poly version of the spider:
Still not perfect though. In this model each section of the spider (the legs and arms, abdomen, and thorax) are all different in terms of complexity and the size of the faces. The thorax has large polygons that make it look comparatively simple to the overly complex legs and arms with their small faces. The next version of the spider attempted to correct that so that each part of the spider looked like they came from the same object rather than 3 different objects.
And voilà! The final model of the spider as it stands now.
Here’s one with some proxy textures and some more progress on the eyes which have been adapted into Icosahedrons:
This model has been a long time in coming. Let’s hope it sticks this time!