Another day goes by another JF gets skipped. What can I say? I can say that school is more important than JF. I can say that I had a seminar today and that it went well if only because I spent time on it instead of JF. Had I done JF and/or other homework, I would be in tears right now lamenting my failure of Physics 496. But even then I feel really cheap. Lemme tell you why. I didn't even make this model today. I made it on April 9 and decided today that since I'm artistically impotent tonight that I would show it to y'all. But that's not the bulk of my cheap feeling. My cheap feeling stems from where I got this idea. I got it from FF8. I don't even own a Playstation let alone FF8. I got an Anime music video and a poster of it and that's it. So I stole an idea from something cheap and cheaply stored it until a day after I had skipped three pages in a week. Ugh. If I feel so bad, why is it that people don't feel cheap doing similar and worse? Am I just sensitive? Am I an anarchist idealist?
This here is a physicist's rendering of the cavendish experiment. It and it's predecessors are the reason why nothing has been done on JF.
Two misses in a week, I'm bad. Sorry, I am giving in to pressure. I am not even getting my work done. I'm just really irresponsible these days. It seems that whenever I get really inspired, I slack off seriously. They don't call me Java n tea for nothing. If you don't get the metaphor, I am like Coffee and Tea because I'm a stimulant and depressant at the same time. When I'm really full of energy, I waste it and when I'm really calm, I'm really depressed. It works out so that I get absolutely nothing done, but I look busy all the time. I can't say that my life has been a complete waste, but it seems that way from time to time. A bachelor's degree in physics? Who cares? Computer programming? To what end? Money? What money? If I did, I'd be more of a failure. 165 pages of JF? What good does it do if no one sees it or e-mails me? JF is seriously getting me depressed. I am eager to get it in full swing and done, but I find it hard to imagine anyone reading it even if I ever get past Scene 5. I will definately postpone Page 5 to at soonest until this weekend. It's close, but all the shit I have to do for school is much closer. Note that I'm taking 22 credits at UW. That's supposed to be an 88 hour week. Thankfully, I'm taking two experimental 5-credit VLPA classes. So really I'm taking the normal 12 credits and two sad excuses for a waste of time. Not that the topics in the classes aren't interesting, but it's just too easy for a physicist like me.
Tonight is a lame JF, but last night's was worse. You won't find it using the VCR because it wasn't there. Last night I skipped Making Of JF for the first time in weeks. Of course, the whole last week has been a bunch of pictures full of excuses of why I'm not doing a real Making Of JF. So tonight's is the best since two nights ago. What I'm doing is I'm showing you a little of what I've been thinking recently. You know how JF is all 24-bit, right? It comes right out of the renderer through anti-alias resize and into your retina at as high a color as your computer can output. Well, these pictures are not 24 bit. The top left is lossless compared to 24-bit (aka the same as 24-bit), but there's only five blooming colors. The top right is a funny one. It is actually the same picture but colors changed to fit the uniform safe colors of the web. Interesting how the skin color is sandy/pink/peach. There isn't a yellow skin color. Which means anyone seeing a picture of Yakuzas using an 8-bit monitor will actually see Hawaiian mobsters. ^_^ So what is a guy to do? Diffusion is okay for large pictures, but not smaller ones. As you see below, the B&W diffusion looks terrible seeing the grayscale source to its left. What about 8-bit diffusion? It's rough. What's this I hear about 4-bit diffusion and palleted? Well, 16 colors is small, but that's what anime is about. You see you have, black, white, skin, skin shadow, shirt, shirt shadow, shirt 2, shirt 2 shadow, pants, pants shadow, hair, hair shadow, eyes, eyes shadow, shoes, shoes shadow. That's sixteen. For highlights, you can diffuse white with the material color. Why are there two shirt colors? Who's wears a single color shirt? So that says that 4-bit is possible. Is it preferrable? Well, how about some math? Say you have a picture collection that you want to distribute to the web. You have 800 pictures, 640x480 with 256 colors. That's 234.4 MB plus palettes. Would you believe that PNG compression is a factor of 14 on simple pictures such as this? So then you have 16.7 MB. That's hefty for a lot of downloaders, but it depends on if you want to just zip it or if you want to give it a hard navigation thing with pop-up ads. If your picture has sixteen colors, you can do 4-bit with no diffusion. You get an instant halving of all your bandwidth cost, 8.4 MB. That'd be nice, but it's very unlikely. Why is it unlikely? Simply because most people don't have two shirts, one pants, skin, black, and white. You need background, more than two characters, and the beauty of diffused CG. So you add a bit of diffusion around the borders. It looks good and some people like the old-school feel of a 4-bit diffuse anime picture. I know I sure do. Diffusion is not well liked by PNG. You see, PNG is based on zlib (the same thing as in winzip) which compresses using these pattern recognization things, right? How does it do it? The first thing it does is gets rid of patterns. If you type 123123123123123123123123, it'll type 123x8 and it'll decode it to type 123 8 times. You see how zip just takes something repeated and ditches it? It's linear, so it only works across the row, not down the column pixels. The second thing it does is looks for patterns within the patterns. If you have a box, it will see the pattern in that, and will repeat the row. Random diffusion makes zlib mad. There is supposedly no pattern to randomness, so a lot of random diffusion makes it heavier than the 8-bit. We certainly don't want to decrease quality in order to increase file size. Ordered diffusion is better, but only if it's got a solid pattern. Also, you can't have much of it or it won't get it at all. What zlib does after that is anyone's guess, but it doesn't decrease size very much. You can see how simple an algorhythm zlib is, right? In fact, it's absolutely awful! You see, the original 24-bit 3d CG w/ gradient had a compression of 6.74 under PNG. That's good for a gradient, full of color, and complex to boot, but check out the lossy JPG! It's 18 kb vs the png's 166 kb. A next-to-lossless JPG version is just 32 kb.
