Adapting recorded action sequences to new situations by adjusting them based on matching visual keypoints between scenes.