Hacker News Reader

Comment by Boxxed | Hacker News Reader

I'm not super familiar with this area so I don't follow... Why is animation any more difficult? I would think you could attach the basic 3D shapes to a skeleton the same way you would with polygons.

View on HN

There are lots of reasons you don’t see a lot of SDF skeletal rigging & animation in games. It’s harder because the distance evaluations get much more expensive when you attach a hierarchy of warps and transforms, and there are typically a lot of distance evaluations when doing ray-marching. This project reduces the cost by using a voxel cache, but animated stuff thwarts the caching, so you have to limit the amount of animation. Another reason it’s more difficult to rig & animate SDFs is because you only get a limited set of shapes that have analytic distance functions, or you have primitives and blending and warping that break Lipschitz conditions in your distance field, which is a fancy way of saying it’s easy to break the SDF and there are only limited and expensive ways to fix it. SDFs are much better at representing procedural content than the kind of mesh modeling involved in character animation and rendering.

One possibility, a little backwards maybe, is to produce a discrete SDF from e.g. a mesh, by inserting it in an octree. The caching becomes the SDF itself, basically. This would let rendering be done via the SDF, but other logic could use the mesh (or other spatial data structure).

Or could the engine treat animated objects as traditional meshed objects (both rendering and interactions)? The author says all physics is done with meshes, so such objects could still interact with the game world seemingly easily. I imagine this would be limited to characters and such. I think they would look terrible using interpolation on a fixed grid anyways as a rotation would move the geometry around slightly, making these objects appear "blurry" in motion.

Sampling an implicit function on a grid shifts you to the world of voxel processing, which has its own strengths and weaknesses. Further processing is lossy (like with raster image processing), storage requirements go up, recovering sharp edges is harder...

But isn't this what the author is doing already? That's what I got from the video. SDF is sampled on a sparse grid (only cells that cross the level set 0) and then values are sampled by interpolating on the grid rather than full reevaluation.

Reply on news.ycombinator.com