Forum archive
Code property and a little voxlap
- Voxlap is great but i can't really understand how can someone implement hidden surface removal in a voxel terrain or texture mapping.Also the little cubes of the voxels in models don't look that good when you approach them(Like the lack of texture filtering in software build games).Do you render and the internal part of the objects or models or is it empty like normal polygon models?When you have a platform with millions of voxels why not render a big voxel instead of the millions voxels? Do voxels has to be in the same size?
Does voxels have any major advantage against polygons?
I removed the question for build engine property,cause i read an interview of Ken in 3dRealms site that has the answer to it.Edited at - I think Ken could keep the IP to his engine because of 3Drealm's policy at the time.
- Voxlap uses raycasting (vertical ray coherence), similar to the method used by NovaLogic's Comanche Maximum Overkill. I only render surface voxels. I don't keep track of colors for interior voxels, and it would be silly to render them. Yes, voxels must be the same size. If you want higher resolution, you can create detail sprites, but sprites hurt the frame rate since they must be drawn separately. No, voxels don't have many advantages over polygons. The best one I can think of is: the concept looks attractive in theory. Unfortunately, in practice, in order to achieve proper frame rate and memory requirements, the code usually turns into a mess of hacks and ugly optimizations. See this thread for further technical details about Voxlap: http://jonof.edgenetwork.org/forum/viewtopic.php?t=30
I left Apogee/3D Realms in 1997. I left with shared rights to the Build Engine, although I never acted on it. I own all work done since then. - I wonder if you can make a model with high res voxels that the cubes aren't visible anymore.
What will be the cost of rendering such a model instead of rendering it with medium poly count(~5000 tris) and with normal maps attached to it. - If you want, you can test the idea out using POLY2VOX. Take a high-quality polygonal model and convert it at various resolutions. POLY2VOX can generate KV6 files up to 1024x1024x512 in size (KVX format only supports up to 256x256x255). The best utility for viewing KV6 files is KWALK, which is part of the Voxlap tools.
- Can it convert a whole animation cycle for md3 models?
Md3 has many meshes like head,upper,lower etc how can i see the whole model?I only managed to load one mesh.
Also for md2 i receive an error not found or not supported format and it didn't load the skin.
It seriously needs a command line arg to set the skin filename.
I set the /v1024 for a md2 model (and kv6 format) and it scales it to 0.28 and the model output is 145*131*512 which is low res.
I change it to scale to /s1.0 and i get a flood3dfill stackoverflow error.
I test it a md3 mesh(upper body) and it managed to output 1024*401*453 which is vey good,but in my p3 cpu it runs with less than 10-8fps in 1024*768.
The output reminds me the software renderer of Unreal,it gives that optical illusion.
You must change the upper limit above the 1024 limit.If you doubled it to 2048 ,the cubes will become almost invisible. - The tool needs a little polish but it is impressive.
I wonder if is a way to transfer geometry from build games.
Could it be possible to take a sprite and output a voxel model?Is there a way to find the correct 3d depth of the pixels in a 2d image?That could be very cool.
The lod is strange and it changes even by the camera vangle.
The lighting in the models needs to be more smooth,because it exposes the triangles edges. Can it convert a whole animation cycle for md3 models?
Yes, but POLY2VOX can only take snapshots at user-specified times. See options: /f#, /n#, /r# for more information.Md3 has many meshes ... how can i see the whole model?
POLY2VOX supports MD3 only to the extent of Jonathon Fowler's Build port - basically MD2 with higher vertex resolution. It does not support multiple meshes at this time.Also for md2 ... it didn't load the skin.
You can use MD2TOOL (google "md2tool Jaakko") to overwrite the default skin filename. POLY2VOX uses this for the skin when reading MD2 files./v1024 ... 145*131*512 ... 1024*401*453
Right. The max dimensions in my released executable are 1024x1024x512.
If you want to increase limits, you can edit POLY2VOX.C and increase LMAXDIM, LMAYDIM, LMAZDIM, and FILLBUFSIZ (to avoid the floodfill overflow). If you don't have Visual C installed, it's probably not worth your trouble. I think you already got your answer: 8-10fps at 1024. Conclusion: polygons beats voxels for both detail and speed.I wonder if is a way to transfer geometry from build games ... possible to take a sprite and output a voxel model?
Yes, I wrote something like this for SLABSPRI, the old voxel editor that is included with the Build tools. It can convert a series of 2D rotated views of an object to 3D voxel, but it depends on a lot of things to work well. In most cases it doesn't. You get much higher accuracy by converting from polygonal models using POLY2VOX.- I tested in my Pentium D 3.0 ghz and i get 30 (+-5) fps.
(I use the upper body of q3 model of Ash from EvilDead)
I also have a look at the old comanche simulator.The voxels fits a small scale game like this.
I would like to test the tool with a really high res model (~50000+ tris) with 2048*2048
settings.(If i can find such a free model)