ML Deformer in UE5 allow users to approximate a complex rig, non-linear deformers or any arbitrary deformation by training a Machine Learning model that runs in real-time in UE.

Our approach:  

• Leverages Machine Learning using UE5 Neural Network Inference (NNI) system.
• Training process using a Maya plug-in automatically generates a range-of-motion with random poses that will guarantee the full deformation space is covered and export FBX and Alembic caches for import in UE.
• Can approximate world-space or skinning relative deformations