After a short venture into suggestive contours, I’m back on my voxel raycaster. I’ve been implementing multiple light sources and phong lighting, as well as drasticly improving the way models are loaded – this allows for even bigger voxel sets to be used.
Limiting factor is now my voxelizer itself, which still needs some representation of the whole voxel data set in memory when voxelizing. Moving this to a streaming / divide-and-conquer algorithm would be nice, but I’m not sure it’s where my focus should be at the moment.
Rendering large voxel sets (like a 1024x1024x1024 glob) sure brings out the detail. This is all at interactive framerates (10-15 fps).