JF make 193
JF Nav
JF Nav
Creation Date: 2002-08-11
Here are two of the four things that I've worked on today. I'll explain the two in the picture, I'll give you a little lesson and then I'll explain the other two things that I've been working on. First off, The text is something I did for fun and possibly for use. You probably can guess where I'll put it. I can put it on CD covers, website splash pages, banners, and wherever can hold something that big and currently uses the lame Verdana font. So here it is. I'll probably color it some wild color, too. I tried using emissive, diffuse, and ambient color to make it look cool, but I didn't take a picture of it. How did I make it? By hand. I took boxes and extruded them a bunch. This is the end result. How did I make the "a"? I simply took a box and extruded it upward twice. Then I took the top and the bottom squares of its left side, ne doso? I extruded them to its left. Then I took the vertexes and welded them. Then I took the new left side and extruded it rightward. Then I extruded up, up, and left. It's just like DDR, ne? Okay, the more important thing that I worked on was the driver. The driver is simply DA 14 in a driver position inside the car. I'll make a new skin when I can. Not yet, though. What is cool about this are two things: first, the animation is perfect. First, he turns the car on. You'll find out why the SUV isn't running at the start of Scene 6. He turns the invisible steering wheel to the right, looks right, hits the breaks, and shouts. Then he hits the gas pedal. Then he looks forward. Then he looks left, hits the break, turns the steering wheel left. This will be perfect for both the manga and the anime. For the manga, I'm going to make a subplot of him being the type of guy that GTA3 is based on. He's driving this SUV full of AK-47 wielding maniacs after Jav and he's going to actually get to talk. Not much, but a bit. You see, I've given far too much monologue to Jav and no one has really talked to him except sensei and each scene has a one-liner. JF Final will be stretched out to many, many pages. I've written down in my dev XML all about how I'm going to expand Scene 1 into 5-10 pages instead of 3. You see, I was afraid to expand my manga for some reason and I still am. Part of that is that the software hasn't been perfect. Another problem is that it takes a lot of disk space. I'm a panzy when it comes to disk space, especially with this handicapped computer. You see, my workstation has a 40 GB hard drive. 10 GB was filled with JF/AltSci stuff alone. Ten more were filled with games (each game is a gig these days, sheesh). Ten more were filled with videos of mine. Of course, I have all the videos on CD and DVD, but I feel vunerable when I'm forced to use a CD-ROM. So everything goes on the hard drive. JF is one of those things. 10 GB would fill 16 CDs. I have fifty blank CDRs, but it'd also take about a week non-stop. Also, one wonders what to keep. Of course, I don't have to keep any of the screen capture bitmaps. I could zip them or even leave it to the 3D .asd files and my models. Also, there's the question of keeping source files for Making Of JF files. Should I keep them zipped, on CD or not at all? If I lost it all, I could simply download it from my website (which I did on this most recent crash). I also keep CDs of JF around just in case. But I guess Scene 6 will be the first scene that is longer than it should be. I'll be happy. I'll just take twice as many screen shots as I usually do and use them much less frugal. Then I'll delete all the source once I've gotten the page done. My plan is to have the first page done in one week. It shouldn't be too hard. You see, all I need is that new Jav40f model to open the door, give a huge expression, and run. Then I need the SUV and it's four occupants. If I do one model per day, I'll have it done in five days. No problem.

I told you that I would explain my other two projects. So here is my explanation. I have this idea for a wonderful 3d building modeller. It would act a lot like an architect's tool. You'd get a top view and you'd be able to make rectangles. You'd make an outdoors, then you'd make rooms. You'd connect the rooms and the outdoors with doors. It would take exactly five click and drags and two clicks to make a building with two rooms and two doors. One major thing is that the building would look good and would be optimized for minimum polygon usage. In any normal 3d Modeller, such a building would take about fifty clicks. Then, there'd be elevations that you could make a roof with. You could make windows with one click and one click-and-drag. Stairways, and cabinets are just two clicks. Texturing would be similar to MilkShape 3D. Then it would save directly to a DirectX file. It would also become a portal system. That's a bit more than I'd like to say about that, though.

The next project I've been working on is terrain. I've been thinking seriously about making a video game that uses an entire city. Players would be encouraged to take a taxi, a bus, a car, or a bike around my lovely town of Seattle, WA, USA and would do something undefined. Since I want to include most of urban Seattle including north of my house (U District) and south of the international district, it's a huge place. It has a 5km radius. Then we do the calculations: a block is 75 m, 10,000 m / 75 m = 133 units. That means that our terrain must be 133x133=17689 vertices, 132x132x2=34848 triangles. 34848 triangles can not be done at 30 fps by a GeForce2, let alone a GeForce or a TNT2, etc. So the only way is to reduce the initial. That's easy. Microsoft's MeshView has a function where it will delete the edges that are the smallest to your order. So if I have a 35k triangle mesh and I say that I want a mesh with similar properties with only 10k triangles, it can do it for me. It isn't lossless, though. You lose some of the smaller hills. But wait, another solution is possible: CLOD. CLOD (continuous level of detail) is an algorhithm that does exactly what the previous thing does, but better -- it is real time, it is percievably lossless, and it is far more thorough. You see, in each frame it takes the edge length and compares it to the distance from the camera. If it is very small, the CLOD snaps the edge away. You won't notice it. It does, however, take a lot of CPU. In fact, with 10,000 triangles, it'll be pretty bad. However, I might suggest a possible good thing about this: the less triangles you have, the fewer collision tests you have to do. CLOD info might even be useful to collision tests. Since it has to do with distance, perhaps the data could be used to find all the nearest triangles. It seems possible. That would seriously decrease the CPU usage. You see, a week ago I also came up with a perfect idea for a "accelerated particle-2d terrain collision algorhithm". It simply finds the most north, south, east, and west of a triangles (a few if statements) and tests whether the point is inside that. If it is, it tests the collision. The most likely case is that it is only within the box of one triangle and thus there is no need for a slow algorhithm. There is only 19 if operations per triangle. This is very good for a terrain of only 2000 polygons or so. For 10k polygons, it is pretty bad. If there were 10k polygons, I would not have to even do a collision, I would be able to calculate the triangle from the X and Z values, since the map is just horizontal scanlines. Anyway, you see my idea, right? Well, it shouldn't be too hard. I'll make a full city game yet.

Talking about a full city game, I've been thinking about buildings, cars, streetlights, trees, etc. They seem to be a big problem. You see, if a person is able to see 100 buildings at once and each is 20 triangles, that's 2000 triangles. What if they see 10 cars (300 triangles min.), 10 people (500 triangles min.), and terrain (approx. 1000 triangles using CLOD) they're down to 35 fps with a GeForce2 GTS. They'd probably be very happy seeing that they're seeing the best of the best graphics ever seen on planet earth (note that this is ~ the specs of GTA3). But that's amazing that such a simple scene could bring out complete chaos to a video card that was top of the line one year ago. A GeForce4 could do it at 50 fps, but who has a GeForce4? Not to mention that a GeForce2 GTS can outperform a GeForce4 MX. I mean, how am I going to animate that? How am I going to market it to people who have a GeForce2+? They say that GeForce2 has penetrated the market fully already. I think that's not exactly true, but it doesn't make a lot of sense. Perhaps I can squeeze some triangles out of a TNT2 or maybe I can improve my CLOD. By the way, I don't own a TNT2 anymore. I gave it away to my friend along with a P2-400. Since I'm running on my webserver, I have a slower computer than he does. ^_^ Except that I am using a GeForce2. Hahaha... Doh!

Above, I spoke of a video game where the player would do something undefined. It's not that the game would have nothing to do, but rather that the game type might detract from the idea of terrain. I have two ideas for a city-based game. The first is an RPG where you are an anarchist who has built a homemade tank (ala Dominion Tank Police) and is destroying US Gov't buildings in an attempt to bankrupt the US Gov't. But think of that: you get to drive around a city and shoot artillery at buildings, ripping them to pieces. Isn't that swell? But there'd be serious strategy. First off, there's the police. There's also the SWAT and helicopters. Then there's actually doing damage. How can you make a home built tank that can take out a building? Well, part of the game would be a physics lesson. I've done the math and with a 100 HP moped engine you could drive around in a small tank with light armor and run a compressed-air tank. That tank would be connected to a turret that could be aimed at the building. Then I did the physics of air pressure. I don't have the numbers with me, but I remember a 10 gallon tank filled at 100 psi would put a hole through the strongest glass firing a 10 kg projectile at 34 m/s. It would also take out the wall behind it. In fact, a well-designed projectile could pierce armor or reinforced concrete at that speed. The physics is there, it just needs to be used. Then there's the strategy. Do you wait until the crack of dawn and blast a few holes and run or do you burn the midnight oil and assault the building in the yellow streetlamps for an hour? Or do you want to take a large number of casualties during the day and have every SWAT member in the US after you? Certainly casualties are the last thing on the mind of an anarchist. In fact, destruction is not the correct way for an anarchist to achieve the ideals of freedom. Taking your freedom by force is no better than taking others' freedom by force. But this game would be a lot of fun. Often times when playing GTA3, I would throw my car at a building at 150 MPH. The car would be ruined and it would probably flip over and explode, but the building remains. Darn that police station! ^_^ And why is it that a perfect car accidentally rolling over explodes? It's useful if you're trying to kill a certain marked man who sits in a body cast protected by an army of police, but it's very unrealistic.

Oh, a month or so ago, I was looking through my JF data XML and I found that each time I save, I am saving the date in an attribute. It looks something like this <Page PID="131" PDate="2002-04-09" PHits="0" LHits="0" RHits="0" Height="185" Width="200" OmitEnd="off"> Would it be cool if I just put those dates into the page somewhere? People could see the date and say, "Oh, Javantea was on a sugar binge on 2002-04-09. Which is last April 9th, six days before his income tax was due." Of course, it's not a created-on date, it's a last-updated date. That means that a bunch of the first ones will be off by a month because I created Making Of JF a month after I started the JF comic, so I had to filter out the comic from the Making Of JF and save them in this different scene called "make". You see? Perhaps you don't, but it's a really cool system. It's all XML and it is easy as heck to edit using my web server. I'm thinking of porting it over to PHP, but don't tell Microsoft that. I've been using MSFT products so long that my brain is only half left. Thankfully, it's the half that thinks in 4D (spatial 3D and temporal time dimension). That's why I've become such a good animator. And people say that good physicists aren't good artists. Anyway, I've been thinking, I should at least set up my server as a Linux box. Not only is Linux faster and better, but it's also free plus distrobution cost. So what is my plan for Linux? Well, I think that I'll first convert all my ASP websites into PHP or perhaps I could make a c++ dll to work with a PHP page. Then when I have a job, I can buy a new motherboard and CPU for my workstation. Then my backup computer will be free for use as as a server. Then I'll install Apache on my Windows 2000 Professional "server". It's a common misconception that Apache is only for Linux. It's actually got binaries for win32. I'll get it to work and then I'll reformat my server's hard drive and install Linux. I'm still looking for a good distrobution of it. I've heard that Mandrake is good. RedHat, Debian, and Slackware sound good because they're so popular. Anyway, I'll set up my server to do my bidding. But I don't want to turn it on once per day just to compile the files. Hmm.. Maybe I could just make a c++ program that inputs my XML and outputs my HTMs that works with DJGPP..

I'd eventually have to learn OpenGL (perhaps I'll be in time for OpenGL 2.0) and then convert my workstation over to Linux. That would work.

JF Nav
Home Characters Making Of Technical Mail News Links |< First < Prev Next > Latest >|  bandwidth version Goto Scene