Test on character consistency. This is intended for use along style loras. This has been trained out of a single AI image trying to improve the training set with a mix of frames from WAN videos for high consistency and different expressions and Flux Kontext for increased diversification on the images of the training set.