NodeIcon Replace

Takes two models in the same topology and seamlessly replaces a part of one model with a part of the other. For example you can take an ear shape of one character and apply it to another character as long as their topology is the same.



The node only affects vertex positions and does not modify topology or vertex order.

The region of replacement is defined by a polygon selection. The replaced polygons from the source geometry are deformed in as rigid as possible manner to seamlessly match the shape of the target geometry. The shape of the target geometry (except for the replaced region) remains unchanged. Thus swapping the source and the target models and inverting the polygon selection will give a different result.

The algorithm is sensitive to initial alignment of the models by rotation and scale and not sensitive to translation.


It is recommended to pre-align the models beforehand. In case when good alignment can’t be achieved increase the number of iterations of the algorithm.

Common Use-cases

  • combine two wrapped scans. For example combine facial expression from one scan with a body posture of the other;

  • fix seams after applying Subset and ApplySubset node;

  • revert changes from previous versions of the character or swap parts between characters done in the same topology;


source geometry

Geometry A geometry to copy polygons from

target geometry

Geometry A geometry to paste polygons to. Should have the same topology as the source geometry

polygon mask

PolygonSelection A set of polygons to be replaced. The polygons can be selected either on the source geometry or the target geometry.


Geometry A target geometry on which the selected polygons were seamlessly replaced with the corresponding polygons on the source geometry. The output geometry keeps visual properties of the target geometry.


seamless fitting

if not set, the basic algorithm is used. It copies vertex positions of selected polygons from the source geometry and pastes them to the target geometry.


number of iterations of the algorithm. Should be increased if the source and the target model parts are not aligned.