First version, later the training is improved in future versions.
Try this prompt "a portrait (gillianxx:1.2) with a UFO behind"
gillianxx-1e.safetensors