Our model can be trained on only a few images (e.g. 10) of a given character labeled with user-chosen keypoints. The resulting model can be used to animate the character on which it was trained by interpolating between its poses specified by their keypoints. We can also repose characters by simply moving the keypoints into the desired positions. To train the model all we need are few images depicting the character in diverse poses from the same viewpoint, keypoints, a file that describes how the keypoints are connected (the characters skeleton) and which keypoints lie in the same layer.