There are two version.
v1 was trained on 33 random images for 1800 steps
Inara Serra version was trained on images from Firefly
Trained on screencaps from Firefly.