Earlier AI techniques could convert an image of one animal into another with a similar pose, but only on the domains it was trained on. However, this new AI is able to take an input image of an animal pose and a style image of an animal it has never seen before and create entirely new images that show the new animal in the input pose.
"COCO-FUNIT: Few-Shot Unsupervised Image Translation with a Content Conditioned Style Encoder" by Kuniaki Saito, Kate Saenko, and Ming-Yu Liu:
https://nvlabs.github.io/COCO-FUNIT/
