Just thought I’d share the results of my first test capturing prim geometry data from the Second Life video stream. If you don’t know what I’m talking about, check out my earlier post (Link).
The above is something I modeled inside Second Life which I was hoping to someday fabricate; a kind of light sculpture. I’d done sketches years ago and while browsing through some sketchbooks last year, rediscovered it. Doesn’t look like much since it’s just raw triangle data, but it imported into Maya cleanly (better than the rocket parts) and I don’t have any doubt I could take that into Pro/ENGINEER.
Now for the bad news. So far I’ve not succeeded at capturing a model from Doom3. I suspect it’s the Vertex Programming that has apparently caused some other problems with users attempting to capture geometry, but until I spend more time with it, I won’t know if the solution suggested works for me. However, this tool is grabbing the normalmaps, which is excellent. I’ve not checked to see if it’s also logging the uv map needed to allow someone to properly associate the normalmap with the geometry file (I have a bad feeling about that actually), but hopefully it does have something. More later.
I don’t have DOOM 3 on my computer, but from what I hear no one else has been able to capture from it either. Quake4 has worked to a limited extent (aren’t they the same engine?)
Currently I’m not grabbing the texture maps’ [U,V], but that’s one of the top priorities in the ToDo list: http://ogle.eyebeamresearch.org/limits#todo. I would love to get some help on this from the small community that already seems to be forming around the software.
I’m pretty sure that the .obj file format allows for texture coord [U,V]’s, and to reference image files on disk (which are captured by GLIntercept actually, not OGLE) so it should be doable. Lazyweb, where are you?!?!?
Yes, both Doom3 and Quake4 use the same engine. Perhaps its the way that Quake4 models are made and animated. I believe they’re segmented similar to Quake3 models. That would mean that the animation isn’t a single rigged mesh like in Doom3 (iirc), but is a group of meshes (head, torso, legs, hands) controlled via “tags” on the mesh segments.
I’ll look into that.
it’s a single mesh, still skeleton based.
the models and textures are in .pak files, open with winzip, why the need to make an app to rip them?
Thanks, CheapAlert. I was just getting ready to check. So much for that.
NoBody, this is only an exercise. Similarly, I could easily have ripped the data for the Quake3 rocket launcher I ripped last year (and say as much in that original post), but the test is for those games and future apps where the data never resides locally. For example, I can go into Second Life and vidrip someone else’s virtual product; even take it all the way to manufacturing. Now, if Second Life had normalmaps…
Doom3 models captured. I’ve documented the issues I had and the solution on the SourceForge OGLE forum.
csven, are you in NYC? Jeffrey Paffendorf (sp?) of SLSalons is coming by Eyebeam on Tuesday to talk about this stuff. I think you’re down with him, so if you’re in the area you would be invited to come as well…
About two and half hours out. And, yeah, I know that avatar.
Check your email.