Indeed I have heard of surface nets before, as well as dual contouring (which I personally consider the “best” complexity/quality tradeoff) But for the time being marching cubes is pretty good, not to mention fast with room for optimization. Especially after using a remesh+smooth modifier it’s decent.
But I’d love to experiment with other meshing algorithms in the future.
Ductaped the Conjure scene description into my meshing prototype to try out a real asset
And then did some manual retopology ontop to check the workflow.
Being able to separate the design phase from the modelling/topology phase is delightful
Almost done integrating the meshing system into Conjure. (no more slow .obj file conversion to get it into Blender) Still a lot of things to fix/add but getting there.
Spent a lot of time optimizing the conversion to mesh. Now it’s pretty fast so I can finalize the rest of the integration with Conjure
This is nearly 6 million triangles meshed on the GPU. Takes about 10 minutes (including time spent doing some manual work, which will be automated later)
Only discovered this today, thanks to a video by Chris Holt. Looks incredible! Will it work on macOS and with Metal on the GPU? Or is it all platform-agnostic?
I’d be really surprised if those numbers are still valid, given the way the market has evolved. However, I guess we’ll have to wait for a Vulkan version then
Maybe the Blender Foundation could give you a hand with some data and/or the Vulkan stuff? This feels like a major update to Blender which should be part of the app itself!
I wish someone would start creating a Metal-based SDF modeling tool for macOS, similar to MagicaCSG. There’s Unbound (currently in closed beta), but that’s an SDF-based game engine in which modeling is only part of the package.
Indeed as Lamia quoted, MacOS is currently not supported.
Metal is a lower level API (and so is Vulkan) Which would just complicate development a lot at this stage. Add to this that I don’t have a MacOS device and the userbase is much smaller. It’s just not worth spending time on right now. But I’d love to look into simpler ways to support MacOS in the future (6-12 months from now)
There is more recent user data, as Francis showed.
And unsurprisingly, not much has changed. Mac still represents a small (but not insignificant of course) part of the Blender users. Which is why I do want to support Mac eventually, just not right now.
It’s all python/glsl at the moment so no build/compile needed. Just a standard blender addon like any other.
By my knowledge, only MacOS has dropped support for OpenGL, so I think Linux should be fine? Just needs to support OpenGL 4.3 at least. (4.3 is needed for the conversion to mesh)