The Tree model went through just as many changes and false starts as the spider model. The circuitous journey begins in 3…2…1…
Always start with reference!
The tree in my animation, Spider, is a tree that has been long dead in a desolate wasteland. The environment is flat, cracked earth with nothing of interest for miles but this one isolated landmark. The spider should be nesting in this still standing fossil and taking shelter from the cold each night in a knot in the bark. The tree model has to be striking from a distance in silhouette, look good in extreme close-up (spider POV), allow for interesting camera angles throughout the animation, and have a nice place for the web and the spider hole.
But, hey, modeling a tree is hard work, and this animation isn’t supposed to take forever and a half to complete. It’s just a layout exercise, right? Enter the Proxy Tree:
Damn, that is a sweet tree you might say- WRONG!
This is the proxy tree that I built using the Maya Paint Effects tool. For those of you who aren’t familiar with Maya Paint Effects, it’s a tool that allows the user to quickly create procedural geometry. It’s especially useful for organic objects like trees, shrubs, plants, etc, because these objects can be particularly time consuming to create by hand. A lot of research has gone into making procedurally generated trees and plants that appear to have grown in a natural way, branching off at random points, arching, twisting, and all that jazz. I’m don’t think much of that research made its way into the Maya Paint Effects system, at least not into the stuff it comes pre-packaged with. The Proxy Tree was built by modifying a tree that was generated by the paint effects tool, merging vertices, cleaning the geometry up, and adding the spider hole and web, but you can see that the underlying structure of the tree is very basic. But that was ok because, at the time, Spider was meant to be nothing more than an exercise in visual storytelling and cinematography.
When I showed this tree to my professor, Ergun Akleman, his comment to me was, “It looks very L-system-y”.
L-systems, or Lindenmayer system is a system of recursive rules that make it easy to describe and model fractal-like objects, such as those found in nature. It’s easy to simplify a tree down to a series decisions about when and where to create a new smaller branch off of a base, but the results do have a certain procedural look to them that can be obvious if you know what you’re looking for. L-system trees are probably great for background objects or dense forrest scenes where the objects in question are not the focus of the story. That’s probably exactly what the Paint Effects tool was built for in the first place, actually: creating a whole lot of complex, but rule-based geometry to fill a background. Unfortunately my animation, Spider, only has one tree and since we spend the entire story in and around the tree it has to look pretty good from nearly every angle.
Not only that, but the tree did not fit in well with my new Low-Poly aesthetic. So I called in Brian who at the time was still working on remodeling the spider as well.
Brian did an excellent job translating his ‘shattered glass’ style to a new tree model. If you’re interested in knowing more about Brian’s aesthetics and how he achieves this look, check out the Spider Modeling Progression post. I particularly like how the hardened edges accentuate the geometry.
Good, good, so we’re done, then right? I even cut a little hole in there and rebuilt the proxy spider web. That means we can go back to cinematography, right? Well…no. First we have to remake the tree again because we’ve made a switch from dirty lo-poly art to clean lo-poly, remember?
Ok, so here’s some shots of the process of converting this Brian Tree to this new Clean Lo-Poly Brian Tree.
First I remodeled the tree using regular box modeling techniques, working with quads. Then I triangulated the model and smoothed it out using the sculpt geometry tool in Maya. Pretty basic stuff.
Here’s some shots of the finished Brian Lo-Po Tree:
OK, so we’re done now, RIGHT?? Because, remember, I’m not a great modeler, so my real objective is to not spend much time on this at all. Remember when this was supposed to take, like, 4 months? Man, those were good times, guys…
Modeling, more than any other part of the 3D-animation pipeline, is easy to get bogged down in. The question is never, “When are you done.” It always ends up being, “When do you absolutely have to stop.” I suppose every department feels this way, but I think inexperienced 3D artists reach this point quicker in the modeling phase than in the lighting or animation phases for instance. We’re used to seeing 3D objects, so we’re better at judging them and knowing when they could be made better, whereas most 3D artists just starting out haven’t developed an eye for good and bad animation, lighting, textures, etc. As a result, modeling tends to eat up a lot of time because of perfectionism, basically.
In my case I wasn’t satisfied with the overall look of the tree. Brian had modeled a dead tree, as per my instructions, but he went a bit further by imagining a tree that had been broken in two, perhaps by a bolt of lighting. The branches come off from the central trunk, but end abruptly and stunted. It makes for an interesting model, but something told me that it was a little too regular. It felt like a human was trying to be random with it, rather than it actually being random. Is that vague enough for you?
Finally I started doing what I probably should have been doing since the very beginning. Picking a tree that I liked, and modeling it from reference. Unfortunately for me I only had one image of this tree, even if it was a fairly good, nearly orthographic image. It took a long while to model it branch by branch, but it’s fun looking back at the progression:
As with the Lo-Po Brian Tree, I started modeling in quads and extrusions before triangulating and smoothing for the final version.
Here’s a few shots of all the model versions all together!
Here’s some quick shots of the final tree model and one showing the difference between the final, triangulated lo-poly version of the tree and the version that’s still made out of quads:
And here’s some shots of the final tree in all it’s glory:
Professor Akleman hasn’t seen the final model yet, so I’m sure that this will still be going through a few changes before I finishing the animation. However, this tree certainly accomplishes many of the original objectives laid out at the beginning of this post. I even tested out the tree with the new spider model to see how they look together:
Whining about modeling aside, I think that this new tree is a major stop forward for the project. I hope that this new tree serves me well in the future!