import torch
from diffusers import DiffusionPipeline
# switch to "mps" for apple devices
pipe = DiffusionPipeline.from_pretrained("stabilityai/stable-diffusion-xl-base-1.0", dtype=torch.bfloat16, device_map="cuda")
pipe.load_lora_weights("ostris/ikea-instructions-lora-sdxl")
prompt = "where is waldo "
image = pipe(prompt).images[0]Ikea Instructions - LoRA - SDXL
where is waldo
No trigger word needed. Weight of 1.0 works well on the SDXL 1.0 base. Negatives are usually not needed, but "blurry" and "low quality" seem to help. You can use simple prompts such as "hamburger" or describe steps you want it to show. SDXL does a pretty good job of figuring out the steps to make it.
Image examples for the model:
sleep
hamburger,, lettuce, mayo, lettuce, no tomato
barbie and ken
back to the future
the dude, form the movie the big lebowski, drinking, rug wet, bowling ball
hippie
fat man, eat pizza, eat hamburgers, drink beer
dayman, fighter of the night man, champion of the sun
nightman, pay troll toll to get into that boys hole
- Downloads last month
- 652
Model tree for ostris/ikea-instructions-lora-sdxl
Base model
stabilityai/stable-diffusion-xl-base-1.0