Experimental LORA trained on mixed dataset mostly anime. Most likely need more training but I like the small changes it makes on low strength.

Recommented on 0.5 strength.

Samples made at 1024 x 1536 resolution 12 steps

Any feedbacks are welcome.