Is there any (semi-)automatic way to generate a subdivision surface (i.e. a low-resolution mesh with subdivision applied) from a high-poly mesh?

For example, say I have a sphere with 1k polys. (Not nice, neatly arranged polys, but the sort of mess you might get from a photogrammetry reconstruction.) The goal is to get a mesh that, when C-C subdivision is applied (with sufficient steps), closely approximates the original mesh.

I'm hoping for something a bit more goal-oriented than applying a ton of decimation to the original and then offsetting it. The more initially automated, the better, but I could live with something that requires an initial SDS and fiddles with it to more closely match a target mesh.

And no, decimate's "un-subdivide" mode isn't doing it. I suspect what I really need is something that takes an initial SDS and fiddles with the vertices to minimize the overall deviation from the target surface.

  • $\begingroup$ did you try the decimate modifier? $\endgroup$
    – Chris
    Aug 24, 2022 at 6:05
  • $\begingroup$ On one mesh, it effectively did nothing. On another, it sort-of worked for about one iteration before turning the mesh into a complete mess. And it only seems to work if the input mesh has nice, neat faces, like you'd get from something that was originally subdivided. $\endgroup$
    – Matthew
    Aug 25, 2022 at 2:02
  • $\begingroup$ I think the closest thing you can get to is the Remesh feature under Object Data Properties $\endgroup$ Aug 25, 2022 at 6:05
  • 1
    $\begingroup$ Remeshing in Blender or through third-party tools like Instant Mesh or more modern tools in the same space, would be what you're looking for. $\endgroup$
    – Robert Gützkow
    Aug 25, 2022 at 14:06
  • $\begingroup$ @HarryMcKenzie, Remesh is... really not what I'm looking for, either. Some of the ideas there are "right", and it might be useful for generating an initial SDS, but really looking for something that will "optimize" an SDS for best fit to an existing mesh. $\endgroup$
    – Matthew
    Aug 26, 2022 at 5:09


You must log in to answer this question.

Browse other questions tagged .