diffusers/docs/source/en/using-diffusers/dreambooth.md

DreamBooth

DreamBooth personalizes a pretrained model to a specific subject from a few images (for example, your cat) by fine-tuning the full weights and binding that subject to a unique identifier in the prompt (sks cat). You can then generate the subject in new settings, lighting, poses, and styles.

DreamBooth checkpoints are typically a few GBs because they contain the full model weights. Load them with [~DiffusionPipeline.from_pretrained] and include the unique identifier in the prompt to trigger generation.

import torch
from diffusers import AutoPipelineForText2Image

pipeline = AutoPipelineForText2Image.from_pretrained(
    "sd-dreambooth-library/herge-style",
    dtype=torch.float16
).to("cuda")  # or "mps", "xpu", "cpu"
prompt = "A cute sks herge_style brown bear eating a slice of pizza, stunning color scheme, masterpiece, illustration"
pipeline(prompt).images[0]

To train your own checkpoint, see Train DreamBooth.