Tyler Britton

pixelninja
For something like this I would recommend making your life easier by keeping the topology consistent across variations. Then you don't need to worry about the UVs at all.

Failing that you could try slicing your meshes in half along the XZ plane with a clip SOP. That would give clean edges to your UV shells and might make them easier to match/transfer.

Thank you. Yes the topology is consistent, it is just that the topology changes between the variations are causing the issues mentioned above. Thank you for the other idea, I will try that out.

I have a base "pebble" geometry that I want to them make variations for. I UV'ed the base pebble, then want to apply those UV's to the geometry variations so that the textures still apply. I want to have lots of geometry variations (2k+) without having to UV and retexture each one.
I have two methods of transferring the UVs, one that moves the variation's geometry to that of the base geometry before UV transfer before switching it back, and one that does not.


The former (let's call that Former) I get some obvious UV stretching due to the variation's geometry being to different than the base's geometry.


The latter (let's call that Latter) does not have this stretching but has a noticeably different profile for the UV shells which also doesn't look good when the texture is applied.


My solution is to see if I can use both, using the Former's UVs but softly transitioning ti the Latter's UVs close to the UV shell's edges. This would ideally let me have the UV shell profile on the Former while also keeping the distortion fixes of Latter in the middle of the UV shell.
To do this, I would need a vertex or point attribute that has that position's distance from the shell's border in UV space. That way I can transition between the two UVs. Is that something that is possible to do, or is it crazy talk?


Any other ideas are welcome. I included some screenshots along with a hip file.

bender83
Have you sent this as a repro to Sidefx yet? We're just sending ours through now - it'll be good for them to have multiple examples.

Ive also noticed resubmissions don't fix the issue - which suggests the error is encoded.

Yes I just sent one.