app.py

import gradio as gr
import numpy as np
import random
import spaces
from diffusers import DiffusionPipeline, DPMSolverSDEScheduler
import torch
from huggingface_hub import hf_hub_download
from ultralytics import YOLO
from PIL import Image
import cv2

device = "cuda" if torch.cuda.is_available() else "cpu"
model_repo_id = "John6666/wai-ani-nsfw-ponyxl-v8-sdxl"
adetailer_model_id = "Bingsu/adetailer"  # Your ADetailer model

# Load the YOLO model for face detection
yolo_model_path = hf_hub_download(adetailer_model_id, "face_yolov8n.pt")
yolo_model = YOLO(yolo_model_path)

if torch.cuda.is_available():
    torch_dtype = torch.float16
else:
    torch_dtype = torch.float32

pipe = DiffusionPipeline.from_pretrained(model_repo_id, torch_dtype=torch_dtype)
pipe.scheduler = DPMSolverSDEScheduler.from_config(pipe.scheduler.config, algorithm_type="dpmsolver++", solver_order=2, use_karras_sigmas=True)
pipe = pipe.to(device)

MAX_SEED = np.iinfo(np.int32).max
MAX_IMAGE_SIZE = 1024

def correct_anime_face(image):
    # Convert to OpenCV format
    img = np.array(image)
    img = cv2.cvtColor(img, cv2.COLOR_RGB2BGR)
    
    # Detect faces
    results = yolo_model(img)
    
    for detection in results[0].boxes:
        x1, y1, x2, y2 = map(int, detection.xyxy[0].tolist())
        
        # Crop the face region
        face = img[y1:y2, x1:x2]
        face_pil = Image.fromarray(cv2.cvtColor(face, cv2.COLOR_BGR2RGB))
        
        # Prompt for the correction model
        prompt = "Enhance this anime character's face, fix eyes and make features more vivid."
        
        # Process the face with the anime correction model
        corrected_face = pipe(prompt=prompt, image=face_pil).images[0]  # Replace with your correction model
        
        # Place the corrected face back into the original image
        img[y1:y2, x1:x2] = np.array(corrected_face)
    
    # Convert back to PIL
    final_image = Image.fromarray(cv2.cvtColor(img, cv2.COLOR_BGR2RGB))
    return final_image

@spaces.GPU
def infer(prompt, negative_prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps, progress=gr.Progress(track_tqdm=True)):
    if randomize_seed:
        seed = random.randint(0, MAX_SEED)
        
    generator = torch.Generator().manual_seed(seed)
    
    image = pipe(
        prompt=prompt, 
        negative_prompt=negative_prompt,
        guidance_scale=guidance_scale, 
        num_inference_steps=num_inference_steps, 
        width=width, 
        height=height,
        generator=generator
    ).images[0] 
    
    # Correct anime face in the generated image
    corrected_image = correct_anime_face(image)
    
    return corrected_image, seed

examples = [
    "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k",
    "An astronaut riding a green horse",
    "A delicious ceviche cheesecake slice",
]

css = """
#col-container {
    margin: 0 auto;
    max-width: 640px;
}
"""

with gr.Blocks(css=css) as demo:
    with gr.Column(elem_id="col-container"):
        gr.Markdown("# Text-to-Image Gradio Template")
        
        with gr.Row():
            prompt = gr.Text(label="Prompt", show_label=False, max_lines=1, placeholder="Enter your prompt", container=False)
            run_button = gr.Button("Run", scale=0)
        
        result = gr.Image(label="Result", show_label=False)

        with gr.Accordion("Advanced Settings", open=False):
            negative_prompt = gr.Text(label="Negative prompt", max_lines=1, placeholder="Enter a negative prompt", visible=False)
            seed = gr.Slider(label="Seed", minimum=0, maximum=MAX_SEED, step=1, value=0)
            randomize_seed = gr.Checkbox(label="Randomize seed", value=True)
            
            with gr.Row():
                width = gr.Slider(label="Width", minimum=256, maximum=MAX_IMAGE_SIZE, step=32, value=1024)
                height = gr.Slider(label="Height", minimum=256, maximum=MAX_IMAGE_SIZE, step=32, value=1024)
                
            with gr.Row():
                guidance_scale = gr.Slider(label="Guidance scale", minimum=0.0, maximum=10.0, step=0.1, value=0.0)
                num_inference_steps = gr.Slider(label="Number of inference steps", minimum=1, maximum=50, step=1, value=2)
        
        gr.Examples(examples=examples, inputs=[prompt])
    
    gr.on(triggers=[run_button.click, prompt.submit],
          fn=infer,
          inputs=[prompt, negative_prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps],
          outputs=[result, seed])

demo.queue().launch()