Print view · Auto-Rig · 1 pageChoose "Save as PDF" as the destination in the print dialog.
Back to the docs

Mesh Morpher documentation

Auto-Rig

1 page · exported 2026-09-06 · https://meshmorpher.com/docs/studio-and-graph/ai/auto-rig

Studio and Graph › AI

Auto-Rig

Predicts a skeleton for a character mesh that arrived without one, using the UniRig model, and writes the bones onto the output mesh: names, parents and bind poses.

Auto-Rig node

The Auto-Rig node

When to use it

Predicts a skeleton for a character mesh that arrived without one, using the UniRig model, and writes the bones onto the output mesh: names, parents and bind poses.

It emits the skeleton and nothing else. The mesh still has no skin weights when it leaves this node, so it will not deform if you pose it. Auto-Skin is the second half of the pair.

The usual chain is Mesh, Auto-Rig, Auto-Skin: skeleton first, then the weights that bind to it.

Pins

Pin Type Description
Mesh Mesh The character mesh you want a skeleton for.
Pin Type Description
Mesh Mesh The same mesh, now carrying the predicted bones. No skin weights yet.

Settings

Setting Type Description
Backend Choice Where the model runs. DirectML uses your GPU and is far quicker; it drops back to the CPU on its own when the GPU cannot take it.
Seed Whole number The number behind every random choice in the prediction. Same number, same skeleton. Change it and you get a different one to compare against.
Sample True/False Let the model wander instead of always taking its safest guess. Off gives you the same tidy, symmetric skeleton every time. On finds extra joints now and then, and also gives you a stray bone in one arm that has no twin in the other. Re-roll with Seed if that happens.
Temperature Number How far the model is allowed to wander when placing joints. Higher is more adventurous and less predictable.
Top K Whole number Only consider this many of the model's best guesses at each step. 0 lets it consider all of them.
Top P Number Cuts off the long tail of unlikely guesses. Lower keeps the choice tighter and the skeleton closer to the safe answer.
Repetition Penalty Number Discourages the model from reusing joint positions it has already placed. 1 leaves it alone.
Candidates Whole number Predict the skeleton this many times and keep the one with the most joints, which in practice is the one where eyes and finger tips survived. Each extra attempt costs another full run. 1 is the quick pass.