Various customer-requested updates and improvements for riggers and animators.
The 2026 release features updates to existing features to make your workflows more efficient.
ML Deformer improvements
A number of improvements have been made to the ML Deformer. These include:
- Improved shape training: When training the deformer, a new Tune Principal shapes during training mode is available, which parameterizes principal shapes so they can be adjusted during training. This can improve results, especially when all the training poses don't fit in the SVD analysis.
- Learn Surface: A new Learn Surface option has been added to the training window, which can improve noisy results.
- Dynamic loading now works with all training options.
- Customizable shape analysis: You can now use Shape Analysis to customize how to train on principal shapes when the amount of pose data exceeds the amount of allocated memory.
- You can now allocate a specific amount of system memory for training pose data in the ML Deformer preferences.
- modelPath has been removed from the metadata, and exportSurfaceInformation, surfaceFrameMap, surfaceRepresentation, surfaceFrameSize, usesPrincipalShapes, shapeConstructionOption, and modelDefinition have been added.
See ML Deformer.
Mesh compare
You can now use Apply Mesh Compare to visualize the difference between two meshes using a heat map. This is especially useful when combined with the ML Deformer, since it can either show where deviations exist between the source and target mesh, or areas where the deformer is performing the most work.
