Hey all, what do we got here? Nothing less than a perfect lesson ripe for the teaching. I started to texture this beast, right? The first thing a person is supposed to do is to split up the groups of textured parts. For a car, it is smart to do the side, the top, the front, and the back. So I did that and I spent a lot of time getting it exact to size and stuff. Silly me when I got the top picture as a result. But it's not magic. Low res textures (256x512) will be ugly if you use them improperly. So what about the lower picture? Would you believe me if I told you it was done with boxes? Maybe not. You see, in all the geometry of this picture there isn't a single box. I made sure of that so that it'd look futuristic. There are a few lines that are pependicular and parallel to the ground, but those aren't very impressive either. So I replaced odd looking n-gons with squares and then used the texture editor to pull the texture coordinates to the form of the boxes. It works very well. In fact, I'm going to do that as much as I can. No more messing around with curves or diagonal lines, all straight perpendicular or parallel lines from now on. ^_^ It won't work all the time, but you should rely on it because it gives you about 100x as much pixel resolution.
Today's picture is pretty simple, it didn't take more than an hour. It might have taken less time if I hadn't painstakenly pressed for accuracy in my measurements of my own real-life limbs. I taped feet markers on the wall and used a piece of tape to mark each body part on the wall. Then, I just translated it into a skeleton using MilkShape 3D's wonderful arbitrary movement interface. You see, I just place every joint on the zero and move it upwards N units where N = 10*(feet + inches/12). That gives me a perfect system. From measurements of a person's body, I can make a skeleton. From the skeleton, I can stretch a mesh. I can draw a picture from a photo of a person or from my imagination. From a stretched mesh, texture, and skeleton, I can have an animated character. That's how 3d Animation works. There's a lot of work getting all of those working together, but just look at JF. I think using the body40e model sized to fit different people with different bodytypes, I'll be able to make all of JF. Of course, you've seen the body44 model and you might think that I should ditch the body40. But It takes a lot of work to go from where body44 is to where body40 is. Body40 is where I want to be right now. At least for Jav. The DA model was really good. It worked well with different models scaled very slightly differently with different skins. The Rave Kiddie model worked fairly well for as early as it was. The Jav models have all been less than satisfactory. Really, I always hope for more, but the Jav models have always lagged behind the secondary characters. But you should think about this idea: models ought to be based on reality. While faces are a whole different story, making awkward looking bodies doesn't work. Whether you're making amateur pr0n or manga, it's important to understand what your characters feel like. That's why I like the skeleton system so much. You only rotate the bones. No deformation is possible from the original model. So unlike poor manga, the character looks the same every time through. But getting it right the first time is the important thing. A measurement of a person that looks like your perfect character is the best way to do it. Don't take a picture. Bad idea. Why? Because a camera is a point source that distorts everything except for directly what it's pointing at. That's why the Seattle P-I got away with making a 1.5 foot fish filet knife look like a 1.5 meter samurai sword. They put the camera at such an angle that made it so distorted with the Seattle Chief of Police behind with his arms above his head so that anyone who saw it claimed that it was not the same sword as they saw in real life. 3D projection ont 2d is not a substitute for actually measuring with a ruler. Some people prefer metric, but I only have a good transparent ruler with metric units sandwiched between English units in tenths and sixteenths. *shrug* The lesson for today is to analyze. Use reason, fail-safe methods, and brute force measurement. No sort of algorhithm is going to tell you the right answer for what your character's body will look like. The golden rectangle won't help you here. No book is going to look right when you import it into your 3d program. You'll end up cursing and wasting time. The right way is spend a year on it. If you don't have a year to spend learning stuff, buy the time of someone who already has. If you don't have either, you don't have what you need. I've spent a year and a half and I've developed a 3d engine, learned c++, ported the engine to c++, created a bunch of models, made five scenes of a comic, and made 181 Making Of JF Pages.
It's only 86 kB, so there's no worry about waiting an hour for a 2 second video. Those buttons below it actually do what they say. Tt requires Internet Explorer and Windows Media Player 7. If you don't meet the requirements, click here for the movie in your favorite avi viewer. It's in DivX 5.02 format, as I hope you already know. Click here for the codec.
This picture may not be the discovery of the 20th century, but it is very interesting*. Anyway, this is simple game theory. It took five minutes to do, but it's implications are substantial. The game is such: you have an object. After a certain amount of time, the object turns into three objects (it has two children). A limit is that the children can be placed to the north, south, east or west of the object. The object only reproduces once. The children are objects as well, so after that same amount of time, they reproduce. This continues on until an object cannot have two children. As you see, I've iterated 5 generations. You see that there are two fifth generation children that can only have one child. The rest can have two children. This is a very interesting game because it has a definite end given the limitation that it ends when a child can only have one child. It ends at five generations. So the object model would be something like:
