A method and system that dynamically partitions a 3-D model into multiple regions. The method determines the vertices to be included in each region according to values assigned to an appropriate color channel for each vertex. After partitioning the 3-D model into different regions, the method allows a user to assign different weights to each region to specify a contribution by a driving shape for deforming that region. Dynamic partitioning may be used on multiple color channels to partition the three dimensional model into more than two regions.