Rocket Raccoon SDXL

Hello, AI artists community! I'd like to share what would become my first ever published LoRA to contribute to the availabe resources SDXL has. In this case, my favorite character ever: Rocket Raccoon, from the Guardians of the Galaxy.

This LoRA works well with different resolutions and I made sure it is able to understand well different styles, for that, I used a dataset with various adaptations of the character, but mainly, his MCU version (movies), and comic books version.

Things to consider:

The LoRA is able to generate Rocket from "photorealistic", cinematic scenes, to outlined comic books style. It also understands well. Use "red eyes" in prompt along with "comic book" and you will get a version that is closer to his original design in comics. Otherwise, the LoRA will lean to generate the MCU version. It sometimes makes him bigger than he should be, in that case, putting descriptions like "tall" or "big" in the negative prompt, and "short" in the prompt will do the work. Make sure to put "cartoon" and "chibi" in the negative prompt to prevent extremelly disproportionate generations (not always necesary).

The LoRA could sometimes make appear weird stuff on his head. In some of the images in the dataset he was wearing goggles, and they sometimes leak even if they're not present in the prompt. Put "goggles" in the negative prompt to prevent this. In the positive prompt if you want to make them appear.

It also have a slight tendency to make weird things under his chin. If this happens, put "goatee" in the negative prompt.

LoRA works good with 1 of weight.

Found this useful? Buy me a coffe!