OmniTransfer is a single system that learns from a whole reference video, not just one image, so it can copy how things look (identity and style) and how they move (motion, camera, effects).
KlingAvatar 2.0 is a system that makes long, sharp, lifelike talking-person videos that follow audio, images, and text instructions all at once.
Scone is a new AI method that makes images from instructions while correctly picking the right subject even when many look similar.