starving4rtist

(feel free to move this thread to off-topic...since it's slowly headed in that direction ) I took advantage of a business trip this weekend to pound out the start of an app using the Silverback3D engine: http://markvandeweghe.com/flash/Statuary1.htm

Bringing this thread from the dead once again. I've made it my goal to work on this project for a few hours every night after work, so progress is being made. I apologize again that this project isn't specifically AM related, but I'm planning on integration in the future Some new things you haven't seen yet: Texture mapping 3ds Importer EDIT: Demo of all functionality I've also been working on cleaning up the code for an alpha release to the public, hopefully in a week or sooner. It's not feature-complete yet, but most basic functionality is working. Feel free to look at the Google code page for this project, which has a few (WIP) pages of information about using the code. My main goal was to make it easy for anyone to implement, which I'm hopefully getting closer to. The standard 'spinning model' project can be done with only 10 lines of code So there's still no direct .mdl file support, but I've gotten pretty good results exporting models from AM to a 3ds and loading them into this.

Phong shading: http://www.markvandeweghe.com/flash/engine9.html Simple toon renderer: http://www.markvandeweghe.com/flash/engine10.html

I actually didn't learn about Papervision until I was pretty far into this project. I haven't looked at it as much as I probably should, but it seems to me that my project will end up very similar to it. I think me using Papervision would basically 'void' this project, as I see the two being very similar. (Please correct me if I'm wrong here) I have a long list of features I'm planning to implement, I actually already have texture mapping working, I just haven't optimized the code enough to post an example online. So I guess the main point I'm getting at (in a very roundabout fashion), is that there is no advantage (yet) of using this over Papervision, it's really just a fun learning experience and portfolio piece for me.

Updated a few things...mainly added specular reflections: http://www.markvandeweghe.com/flash/engine6.html Also, I'd like to gauge interest...if I were to develop AM-specific model and animation importers for use with this, how many of you would actually use it? It would take me a fair amount of work to add the file importers, but I'd definitely consider working on it if enough people wanted me to.

My goal at some point is to make model importers for this. Once that is done, the basic code required to load and display a model would be: - Create "3d Engine" object - Create model object and call LoadFromFile() function to import geometry - Load the model into the engine object And then at each frame, you can just call the DrawAll() function of the engine. Before the rendering step, you can make any adjustments (translation/rotation/animation/etc.) to the model that you want. More simply put: yes, this will eventually be able to be used as a hash->flash workflow with very few steps. As far as animation: the older examples that I posted used bones and keyframes exactly like AM does, you can set a bone's rotation for a given frame, and it will automatically interpolate for all frames in-between. Depending on how much time I want to spend working on these features, it's entirely possible to have a system where users can import model AND animation files from AM and display them. FPS is the number in the top left corner. The file is 117kb right now, but much of that size is due to the 66kb texture that I have embedded in the project but not in use right now. I'm guessing most of the rest of the file size is because I'm storing 1500 polygons and vertices directly in the code. Once I write a file importer and use a texture loader, file size will be (IMO) insignificant for most web use.

(un)fortunately I already have another update with significant performance increases. Just like before, I'd be very interested to hear what speeds people are getting with this: http://www.markvandeweghe.com/flash/engine4.html

Here are the results from a fair amount of optimization and code re-working: http://www.markvandeweghe.com/flash/engine3.html There are a few issues with culling at this point, those will be fixed soon. My main goal with this is to get it fast enough to be practical for real-time applications. If you guys are bored, I'd be interested in hearing what sorts of frame rates you're getting for this example as well as the basic specs of the computer you're running it from. Also...sorry that this thread has been constantly straying further and further from AM-related material. I promise it will loop back soon.

I guess I'm confused what you're getting at with this. I'm definitely interested in hearing your ideas, but I'm not sure how this would work. A (simple) triangle generally has 3 colors, one at each vertex. By assigning colors to each vertex, we can make it look like all of the triangles blend together. If there's just a linear 2-point gradient, I'm not sure how that would map onto a triangle? In other news, environment mapping: http://www.markvandeweghe.com/flash/engine2.html It's not 100% accurate yet, this is more of a proof of concept that it's possible to implement. I also did some reading up last night, and I think I'm going to divide this project into 2 parts: a (relatively) fast gouraud shading renderer for real-time manipulation of the models, and a phong shaded renderer for final output when time in as high of demand.

My previous renderer used Flash's vector API to draw each polygon as a triangle. My current goal is/was to create a renderer that allowed smooth shading (vertex normals instead of polygon normals) and texture mapping. I couldn't figure out any good way to draw a triangle with a 3-color gradient between each vertex with any built-in functions. And even if I could do that, texture mapping would still be impossible to do. I settled on doing a per-pixel renderer that goes through most of the same steps that graphics hardware does. Although it's not very efficient (there's a reason we try to use hardware to do these things), it's a step closer to my goal and something I enjoy working on.

Returning once again with an update (I somehow keep coming back to this project): - Re-wrote much of the code to use a new, much more efficient data structure - The new data structure makes importing files much easier (to be implemented later) - Completely re-wrote the rendering code. It now uses per-pixel rendering, which allows for smooth shading (as seen in demo), texture mapping, normal mapping, etc. - Frame rate is far lower that I'd like it to be, but not too bad considering the model is around 1500 polys http://www.markvandeweghe.com/flash/engine1.html I'm considering putting together an on-line modeling/texturing app. Does anyone have any ideas/suggestions for what would be helpful or useful to have online?

Yeah...the physics leave something to be desired. Right now it just has really basic physics (speed up when going down a hill, slow down when going up). The assignment was more about splines than anything, so that's where I put the most work in. If I get some time in a bit I'll definitely look more into the physics.

Ok, I think I've gotten it working now. There's a link to the program in the webpage above, here's a direct link: https://mywebspace.wisc.edu/vandeweghe/web/...MarsCoaster.zip

Something a little different...I hope it's okay for me to show this off here. I just finished writing this program for one of my classes. All of the modeling was done in AM, exported to .3ds, and then imported into my program. I was pretty happy with the way it turned out. Pics and description are here: https://mywebspace.wisc.edu/vandeweghe/web/...case/code5.html I'm working on getting an executable uploaded so you guys can try it out (it looks lots better in motion), but I'm running into problems with Visual Studio. On a different note, did anyone ever release updated exporters to .3ds or something similar? I've done a few tests with textures, and it seems like 1 decal/texture works fine, but once there are 2 or more decals on a model, the exported model has a lot of messed up UV's. Anything I don't know about?