The actor within a video is isolated from the action while our algorithm computes how the environment evolves. After evolution, the prediction of the new environment is generated, overlayed on the original data, while motion is preserved.