Week 9: Coding Surveys, Learning about Texturing
I think there were two primary themes for this, my ninth week at the ArticuLab: 1. analysis of survey results, and 2. learning new techniques in Maya.
At this point, both agents have been re-topologized and are at acceptable polygon counts (10k and 12k). They are Maya-ready. However, there’s probably still a lot of work that remains to be done. The Peer Tutor agents are motivated by factors different from those of models intended for a video game or personal project. As I see it, models for video games and personal projects are complete once they achieve the desired aesthetic quality – that is, once they have the correct anatomical proportions, have enough details, and fit the general “look” of the project. Our agents, on the other hand, might be anatomically correct and look polished – but that hardly means they’re near completion. Instead, because our agents must work as peer tutors, they won’t be done until survey results indicate that kids would want to work with them. It’s a cycle of potentially numerous different iterations.
As such, I have a number of options at this point when it comes to my work on the agents. I described these options to Dave in the following email:
“Based on our discussions from previous weeks, it seems like the models we have now will almost certainly undergo many changes before becoming usable. If I remember correctly, the changes we mentioned were the following:
Change 1: Give the models more African American features (since the current models, even with skin color changed, always look Caucasian)
Change 2: Make the models look older (something that seems even more necessary given our recent survey results)
Both of the aforementioned changes are pretty significant. I see myself as having a few options:
Option 1: Work on changing the models, but work on them in Maya so that no re-topologizing is necessary. Put off rigging and animating unless there's time at the end (dubious).
Options 2: Don't change the models - use the Caucasian-looking ones we have now. Even this second option, however, branches into two "sub-options:"
Option 2A: . Re-color and re-texture the models. Although I can bring our old textures from ZBrush into Maya, they don't look very good on the re-topologized models (since the underlying polygons have been re-arranged). I would need to color and texture the model in Maya, which is a harder task than in ZBrush. It involves creating a UV map of the model, essentially "skinning" the model so that we have a 2D "skin" onto which we can apply color. Other maps (diffuse maps, normal maps, etc) can also be applied to make the model look better.
Option 2B: Don't make changes to the models (or only minimal changes). Don't re-apply textures. Stick with the original plan of rigging and animating.
So, I think a technical perspective of the situation is one that prioritizes continuity of the project: what option maximizes the number of usable assets we have? In that case, I'd suggest Option 1. Option 2 is kind of like taking one step forward and two steps back. Going with Option 2, the models may be rigged/animated or textured, but those animations and the textured models will be of no use to future modelers/animators since they will inevitably need to go back and make changes to the original models. With Option 1, at least some of necessary changes will have been started.
From a personal standpoint, I am fine with all options, though I lean towards Option 2 (particularly 2A). I'm interested in learning how to texture in Maya because it's through textures that low-poly models end up looking so good. Rigging and animating also intrigue me, but they seem like steps that come after texturing. As for project continuity, Option 2 leaves future modelers/animators with two workable models, albeit two very Caucasian and young-looking models.”
We ultimately decided to go with Option 2A, since it would allow me to learn a skill that I’ve been interested in for a while. Accordingly, I spent most of Monday watching tutorials on Lynda.com and YouTube regarding how texturing works in Maya. I am continually impressed by how powerful Maya is as a modeling/animation software! What’s more, it doesn’t “dumb” itself down either – there are quite a few things that the tech-savvy individual can do with MEL scripts.
On Tuesday, I continued working on texturing the models. In addition, I began coding the survey results. We designed the survey so that quite a few of them have quantifiable answers – actually, any question that is based on a Likert scale can be quantified. This means we can assign values to each option in the Likert scale. For instance, the answer to the question, “How well does Character () fit in at school?” could be scored with a 2 (“One of the most popular kids”), 1 (“A little popular”), 0 (“neither popular nor unpopular”), -1 (“A little unpopular”) or a -2 (“One of the least popular kids”). A coding scheme allows for the quick comparison and analysis of data.
Unfortunately, despite the apparently simplicity of the concept, coding takes a lot longer than I would have expected (or maybe I’m just very slow with Excel sheets, who knows). At any rate, those two tasks – coding the survey results and testing some texturing techniques – took up the greater part of the day.
Wednesday followed in much the same way as Tuesday. When the models are brought into Maya, it’s much easier to notice minor distortions in the polygons. So, I spent some time cleaning up the models while watching tutorials. I think that the next modeler/animator will likely have to re-do the models’ eyes so that the textures look better in Maya (they don’t, currently). He or she may also choose to re-do the eyelids, which I added as separate subtools in ZBrush but should really be appended to the face directly.
On Thursday, I began working on creating eyes that were better textured for the peer tutor models. The current eyes, while looking pretty decent in ZBrush, don’t translate well into Maya. I think that ZBrush might have some default lighting and shaders that make most ZBrush models look fairly OK. When I brought the eyes into Maya, however, I noticed that they A. no longer fit perfectly into the eye sockets of the re-topologized model and B. have textures that don’t look very good. So, I found a really nice tutorial on YouTube that detailed how to make a realistic eye here (http://www.youtube.com/watch?v=FUA1pfIE2FQ). Unfortunately, the guy who made the tutorial is clearly very experienced with Maya and often does small maneuvers that I can’t quite follow. I ended up re-watching the first 10 minutes of a 40 minute long tutorial over 7 times just to make sure I was following all the steps correctly.
Worked on creatures.
On Friday, I recorded more survey results and continued making the new eye. I’ve described most of what these two processes entail, so I’ll just say here that overall, it was a productive day! I was able to finish both coding and creating the eye. Next week, I’ll work on wrapping everything up and making sure that all assets for the Peer Tutor models are ready to go for the next modeler/animator.