I followed this guide on the docs up to the point of training the model in UE’s ML Deformer window. It all went smoothly but now the problem is that on the ‘Training Target’ side of the viewport, the animation is not playing, except on the couple of clusters I added (back in Maya). The main bones stay in their frame 0 position (I’m testing this on just an arm so these are the upper arm, forearm, and wrist). The green deltas reflect this, so that the whole arm being in the wrong pose is part of the delta, except on frame 0.
Am I misunderstanding something/did I miss a step?
Possible problems checked:
- All the animated bones are included in the Bone List
- The animation plays correctly in the animseq viewer and ingame
- Fixed the animseq length by not importing the last frame, so there was no longer a warning in the ML Deformer properties panel
- Deleting the base mesh from the fbx didn’t do anything
- The deltas are reflected ingame after going through with the training
UPDATE: My problem was that the target alembic didn’t have any animation in it, and I was mixing up Target with Base. There was no animation because during data generation the Target joints were not moved to match the Base joints like I expected.
TLDR: In Maya you have to constrain the joints of the Target rig to the joints of the Base rig. The Data Generator is going to move just the attributes you put in, and expect both meshes to change.
I wish that were expressed in the documentation, but I guess it makes perfect sense.