This node performs a multi-step quasistatic-style tissue simulation to deform a skin. This multi-step processes controlled and gentle way to optimize simulation stability, quality, and consistency between similar poses.

Armature Deform is used together with Armature Capture. Armature Deform performs the quasistatic simulation and deformation, while Armature Capture prepares the proxy geometry that’s used for simulation.

Armature Deform allows you to achieve organic tissue deformations that look more realistic than just applying Bone Deform SOP. The volume-preservation feature offered by the underlying Otis solver creates nonlinear deformations that are hard to achieve or approximate using linear blend skinning (LBS). The difference between LBS is significant if there is a large gap between the bones and the skin. If the skin fits closely to the armature geometry, there may be less difference between the output of Armature Deform and LBS.

To get faster deformation, only use the amount of detail in the proxy geometry that you need for your skins mesh. This amount of detail is controlled by the Element Size parameter on Armature Capture SOP .

To get good performance when cooking multiple poses using Armature Deform, be aware of the caching behavior of this node. To ensure that Armature Deform can cache its internals such as the capture of the skin relative to the proxy geometry, make sure the Animated Pose input is the only time-dependent input. Having other inputs of Armature Deform time dependent may result in slow cooking times.

Armature Deform internally runs a quasistatic-style tissue simulation. This simulation cannot be treated entirely as a black box. To get good results within a reasonable time, tune the Element Size on Armature Capture large enough so that Proxy Geometry to Simulate has the fewest primitives (tets) as possible. If the proxy geometry has a large number of tets, such as close to a million or more. Your simulation may not succeed with the default settings on Armature Deform. In that case, the output of Armature Deform will appear empty and the geometry spreadsheet will show NAN values for the position attribute P.

If your skin mesh requires highly detailed proxy geometry, there are two settings on Armature Deform you can adjust to stabilize the simulation: Steps and Iterations. As a general rule, if you make the proxy geometry more detailed by dividing Element Size by two, you should multiply Steps by two to compensate. You should also increase the Iterations significantly. The amount is largely dependant on the mesh connectivity.

If your simulation has a detailed proxy geometry, the following suggestion may help:

• Increase the Element Size on Armature Capture to a level where the simulation succeeds.

• Gradually, in a series of steps, decrease the Element Size on Armature Capture.

  • After each step, try increasing Steps and Iterations to compensate which preserves the quality and stability of the deformed end result. For this process, it may be useful to look directly at the deformed proxy geometry on the second output of Armature Deform.

Armature Deform has a variety of potential uses. An example is data-set generation for an ML Deformer. Armature Deform can generate skin deformations based on training poses. Armature Deform relies on the Otis solver’s capability to perform a quasistatic simulation completely independent of frames and time. This allows training poses to process more efficiently than previous Houdini versions where each pose required a time-dependent simulation in DOPs and multiple frames for each pose.

The timeless simulation feature can generate training data in few ways. You can assign a training pose to each frame in a frame range and run Armature Capture over the frame range. Another way would be a for-loop construct in SOPs that runs over all poses.

See Machine Learning documentation for more general information.