pip install torch matplotlib Pillow diffusers transformers open_clip_torch

import PIL
import torch
import requests
import open_clip
from open_clip import SimpleTokenizer
from io import BytesIO
from diffusers import DiffusionPipeline
from transformers import CLIPImageProcessor, CLIPModel


def download_image(url):
    response = requests.get(url)
    return PIL.Image.open(BytesIO(response.content)).convert("RGB")

# Loading additional models
feature_extractor = CLIPImageProcessor.from_pretrained(
    "laion/CLIP-ViT-B-32-laion2B-s34B-b79K"
)
clip_model = CLIPModel.from_pretrained(
    "laion/CLIP-ViT-B-32-laion2B-s34B-b79K", torch_dtype=torch.float16
)
coca_model = open_clip.create_model('coca_ViT-L-14', pretrained='laion2B-s13B-b90k').to('cuda')
coca_model.dtype = torch.float16
coca_transform = open_clip.image_transform(
    coca_model.visual.image_size,
    is_train=False,
    mean=getattr(coca_model.visual, 'image_mean', None),
    std=getattr(coca_model.visual, 'image_std', None),
)
coca_tokenizer = SimpleTokenizer()

# Pipeline creating
mixing_pipeline = DiffusionPipeline.from_pretrained(
    "CompVis/stable-diffusion-v1-4",
    custom_pipeline="clip_guided_images_mixing_stable_diffusion",
    clip_model=clip_model,
    feature_extractor=feature_extractor,
    coca_model=coca_model,
    coca_tokenizer=coca_tokenizer,
    coca_transform=coca_transform,
    torch_dtype=torch.float16,
)
mixing_pipeline.enable_attention_slicing()
mixing_pipeline = mixing_pipeline.to("cuda")

# Pipeline running
generator = torch.Generator(device="cuda").manual_seed(17)

def download_image(url):
    response = requests.get(url)
    return PIL.Image.open(BytesIO(response.content)).convert("RGB")

content_image = download_image("https://huggingface.co/datasets/TheDenk/images_mixing/resolve/main/boromir.jpg")
style_image = download_image("https://huggingface.co/datasets/TheDenk/images_mixing/resolve/main/gigachad.jpg")

pipe_images = mixing_pipeline(
    num_inference_steps=50,
    content_image=content_image,
    style_image=style_image,
    noise_strength=0.65,
    slerp_latent_style_strength=0.9,
    slerp_prompt_style_strength=0.1,
    slerp_clip_image_style_strength=0.1,
    guidance_scale=9.0,
    batch_size=1,
    clip_guidance_scale=100,
    generator=generator,
).images

output_path = "mixed_output.jpg"
pipe_images[0].save(output_path)
print(f"Image saved successfully at {output_path}")