Training continued from v1.0, which was trained on 256p to 1024p. v1.1 was trained at 1024p to 1536p.
Simple LoRA for generating fake album covers.