import torch
from diffusers import DiffusionPipeline
# switch to "mps" for apple devices
pipe = DiffusionPipeline.from_pretrained("Norod78/sd15-muppet-blip", dtype=torch.bfloat16, device_map="cuda")
prompt = "A painting of Cthulhu as kermit the frog, very detailed, clean, high quality, sharp image, based on H.P Lovecraft stories"
image = pipe(prompt).images[0]SDv1.5 sd15-muppet-blip model trained by Norod78 with Huggingface Diffusers train_text_to_image script
For better results, use an explicit name of a muppet such as "Kermit, Cookie monster, etc" or simply use "muppet"
A few sample pictures generated with this mode (more available here):
A painting of the cookie monster, very detailed, clean, high quality, sharp image, based on H.P Lovecraft stories Negative prompt: grainy, blurry, text, watermark, inconsistent, smudged Steps: 32, Sampler: DPM++ 2M Karras, CFG scale: 7.5, Seed: 3320437546, Size: 768x640, Model hash: 9b5251e8, Model: sd15-muppet-blip, Batch size: 4, Batch pos: 2
An oil painting of kermit the frog muppet in flemish baroque style, very detailed, clean, high quality, sharp image, John Philip Falter, Very detailed painting Negative prompt: grainy, blurry, text, watermark, inconsistent, smudged Steps: 20, Sampler: DPM++ 2S a Karras, CFG scale: 7, Seed: 3762044222, Size: 512x512, Model hash: 9b5251e8, Model: sd15-muppet-blip, Batch size: 4, Batch pos: 3
- Downloads last month
- 20