Meat Boy from Super Meat Boy game series. This LoRA was trained using natural language captions. You can write the "Meatboy" tag first to enhance the character tag's impact, and then add it to the prompt body: "Meatboy is a cube-shaped creature…does something.

Main trigger:

Meatboy, cube-shaped creature

Options

grin smile, bandaid on head

May help

Black eyes

Examples of prompts from the training set:

Meatboy. Digital cartoon illustration of a red, cube-shaped, humanoid character with black, oval eyes and a small, black mouth. Meatboy has a single arm raised, pointing upward with its index finger. Its other arm is slightly bent and positioned to the side. The background is a chaotic, red-tinted, broken landscape with scattered, large, rectangular blocks and debris. The character's body is shiny, with highlights on its face and arm. 

Meatboy. This is a colorful digital cartoon illustration featuring a red, cube-shaped character standing on a sandy beach. Meatboy has a simple, blocky design with large, black, oval eyes and an open mouth, giving it an expressive, surprised look. Its arms are small and rounded, positioned at its sides. The background showcases a bright blue sky with a few white, fluffy clouds.

Meatboy. Digital drawing of two cartoonish, plush cube-like creatures. The pink, square-shaped creature with a white flower on its head and large, expressive eyes is smiling and sitting on the back of a red, rectangular creature Meatboy with similar large eyes and a wide, white-toothed grin. The pink creature's arms are resting on the red creature's shoulders. The red creature is running forward with a joyful expression. 

Notes:

1. The drawbacks of this model include the fact that MeatBoy's missing tooth isn't reproduced. The side view will mostly be to the left. The artist styles have little effect on MeatBoy; the training data is mostly cartoon/3D images. Interaction with other characters is limited; characters can carry him, but other interactions will cause him to lose his cube shape.

2. There are many images in the preview where the Meatboy isn't red. This isn't a flaw in LoRA. I wanted to show that the model can easily change its color if you choose.

3. This is the first LoRA I trained for the Anima model. I don't know if it was the character, its simple shape, the fact that the description was in natural language, the model only partially knew what it looked like, or if it's just a very well-trained model, but an acceptable result was already achieved at training step 183. The published version is training step 366; it understands simple 3D/cartoon styles a little better. I was wondering which tagging method would be best: just "danbooru" tags, natural language, or some combination. Ultimately, I chose natural language, with a single "Meatboy" tag at the beginning to help the model better remember the character. Keep in mind that without a description of the character's cuboid body, it won't render correctly.