Rob Chapman

After much further digging its the UV's causing the problem, deleting all UV attributes then problem goes away. so definately its an issue with Unreal engine 5.4 inmport end of the equation. Its a funny old game to get increasing vertices (points) every iteration doing a round trip from houdini > unreal > Houdini > unreal again. different settings with smoothing groups and various import FBX options with creating or not normals and tangents gave varying vertex count results, but this is an Unreal problem. normally(ha get it?) its not an issue for just an exported skeleton mesh to work as is, but in this case I needed the skeleton mesh vertices count to match EXACTLY another exported alembic file from Houdini, to train the ML Deformer correctly

so the training worked in the end ONLY if I made the alembic file from a point deformed version of the skeleton mesh via Houdini > Unreal > Houdini. it doesnt work from the orig Houdini export as the vertices ID number is different also so the deltas were WAY out of whack.

Happy days

omg, trying to train unreal engine skeleton mesh with muscle and tissue sims made in houdini and the round trip FBX Unreal > FBX houdini > FBX Unreal - I get 2 extra vertices (points) added in unreal engine (5.4)
Whilst everything imports OK the ML deformer training data needs the vertices count to match up between the FBX and the ground truth alembic geoemtry cache of the muscle sim...

this is an impossible situation. changing some of the FBX import settings in unreal I get varying amounts of vertices each time (generate normals, compute normals, preserve smoothing groups etc), the only way it seems to work is re-export the FBX from unreal - use this for the pose deformed muscle sim, and DO NOT REIMPORT to use as training data otherwise I get 2 extra verts added. OUCH!

just found this...



with a link to capybara all muscled up correctly with Franken muscles in houdini 20 hip, google drive file link in the description !