app.py

import gradio as gr
import asyncio
import fal_client
from dotenv import load_dotenv
import os
from pathlib import Path
import time
import json

load_dotenv()
os.environ["FAL_KEY"] = os.getenv("FAL_API_KEY")

async def generate_paris_images(product_name: str, image1_path: str, image2_path: str, woman_prompt: str, man_prompt: str, girl_name: str, girl_hair_length: str, girl_hair_style: str, girl_hair_color: str, boy_name: str, boy_hair_length: str, boy_hair_style: str, boy_hair_color: str, batch_size: int, progress=gr.Progress()):
    start_time = time.time()
    print("Progress: 5% - Starting Paris image generation...")
    progress(0.05, desc="Starting Paris image generation...")
    
    # Upload all images in parallel
    upload_tasks = [
        fal_client.upload_file_async(str(image1_path)),
        fal_client.upload_file_async(str(image2_path)),
        fal_client.upload_file_async("template/man_pose.png"),
        fal_client.upload_file_async("template/woman_pose.png"),
        fal_client.upload_file_async("template/woman_clip_mask.png"),
        fal_client.upload_file_async("template/man_clip_mask.png")
    ]
    
    [image1_url, image2_url, man_pose_img, woman_pose_img, woman_clip_mask, man_clip_mask] = await asyncio.gather(*upload_tasks)
    
    print("Progress: 40% - Uploaded all images")
    progress(0.4, desc="Uploaded all images")

    # Replace {hair_feature} placeholders with user hair descriptions
    woman_hair_desc = f"{girl_hair_length} {girl_hair_style} {girl_hair_color} hair,"
    print(f"Final woman hair description: {woman_hair_desc}")

    # Handle bald case for man's hair description
    if boy_hair_length == "Bald":
        man_hair_desc = "bald,"
    else:
        man_hair_desc = f"{boy_hair_length} {boy_hair_style} {boy_hair_color} hair,"
    print(f"Final man hair description: {man_hair_desc}")
    
    woman_prompt = woman_prompt.replace("{hair_feature}", woman_hair_desc)
    man_prompt = man_prompt.replace("{hair_feature}", man_hair_desc)

    print(f"Final woman prompt: {woman_prompt}")
    print(f"Final man prompt: {man_prompt}")

    handler = await fal_client.submit_async(
        "comfy/LVE/paris-couple",
        arguments={
            "loadimage_1": image1_url,
            "loadimage_2": image2_url,
            "loadimage_3": woman_pose_img,
            "loadimage_4": woman_clip_mask,
            "loadimage_5": man_clip_mask,
            "loadimage_6": man_pose_img,
            "woman_prompt": woman_prompt,
            "man_prompt": man_prompt,
            "girl_name": girl_name,
            "boy_name": boy_name,
            "batch_size": batch_size
        }
    )

    print("Progress: 60% - Processing images...")
    progress(0.6, desc="Processing images...")
    
    result = await handler.get()
    print(result)

    end_time = time.time()
    processing_time = end_time - start_time
    print(f"Progress: 100% - Generation completed in {processing_time:.2f} seconds")
    progress(1.0, desc=f"Generation completed in {processing_time:.2f} seconds")
    
    # Fix the URL extraction logic
    image_215 = []
    image_818 = []
    if "outputs" in result:
        if "215" in result["outputs"]:
            image_215 = [img["url"] for img in result["outputs"]["215"]["images"]]
        if "818" in result["outputs"]:
            image_818 = [img["url"] for img in result["outputs"]["818"]["images"]]

    print(f"Image 215: {image_215}")
    print(f"Image 818: {image_818}")

    # Return all generated image URLs and processing time
    # Get the first key from outputs dynamically
    return (
        image_215,
        image_818,
        f"Processing time: {processing_time:.2f} seconds"
    )

def change_product_preview(product_name):
    # Load prompts from JSON file
    with open('prompt.json', 'r') as f:
        prompts = json.load(f)
    
    # Find the matching prompt data
    prompt_data = next((item for item in prompts if item['title'] == product_name), None)
    
    if prompt_data:
        return (
            f"thumbnail/{product_name}.png",
            prompt_data['woman'],
            prompt_data['man']
        )
    return None, "", ""

with gr.Blocks() as demo:
    with gr.Row():
        product_name = gr.Dropdown(label="Product Name", choices=["Winter", "Classy", "Night Out", "Romantic"], value="Winter")
        product_preview = gr.Image(label="Product Preview", type="filepath", value="thumbnail/Winter.png", height=500, width=500)
    with gr.Row():
        image1_input = gr.Image(label="Upload Woman Image", type="filepath", value="user3-f.jpg")
        image2_input = gr.Image(label="Upload Man Image", type="filepath", value="user3-m.jpg")
    
    with gr.Row():
        with gr.Column():
            woman_prompt = gr.Textbox(
                label="Woman Prompt",
                value="Close-up, portrait photo, a woman, {hair_feature} wearing a cream-colored wool coat, chunky knit scarf, and matching earmuffs, standing on the same snow-dusted cobblestone street, illuminated Eiffel Tower glowing golden in the background, snowflakes sparkling in the warm streetlight glow, looking at camera with gentle smile."
            )
            girl_name = gr.Textbox(
                label="Girl Name",
                value="julie delpy"
            )
            girl_hair_length = gr.Dropdown(label="Girl Hair Length", choices=["Short", "Medium", "Long"], value="Long")
            girl_hair_style = gr.Dropdown(label="Girl Hair Style", choices=["Straight", "Wavy", "Curly"], value="Straight")
            girl_hair_color = gr.Dropdown(label="Girl Hair Color", choices=["Blonde", "Brown", "Black", "Brunette", "Redhead", "Bronde"], value="Bronde")
        with gr.Column():
            man_prompt = gr.Textbox(
                label="Man Prompt",
                value="Close-up, portrait photo, a man, {hair_feature} wearing a dark navy wool peacoat, cashmere scarf, and leather gloves, standing on a snow-dusted cobblestone street, illuminated Eiffel Tower in the background glowing golden against the night sky, gentle snowflakes catching the warm glow of vintage streetlamps, looking at camera with confident expression."
            )
            boy_name = gr.Textbox(
                label="Boy Name",
                value="ethan hawke"
            )
            boy_hair_length = gr.Dropdown(label="Boy Hair Length", choices=["Short", "Medium", "Long", "Bald"], value="Short")
            boy_hair_style = gr.Dropdown(label="Boy Hair Style", choices=["None", "Undercut", "Mullet", "French Crop", "Slicked Back", "Fade", "Buzz Cut"], value="Undercut")
            boy_hair_color = gr.Dropdown(label="Boy Hair Color", choices=["None", "Blonde", "Brown", "Black", "Brunette", "Redhead"], value="Black")
    
    batch_size = gr.Slider(minimum=1, maximum=8, value=4, step=1, label="Batch Size")
    
    generate_btn = gr.Button("Generate")

    with gr.Row():
        image_output = gr.Gallery(label="Generated Image Raw")
        image_output_processed = gr.Gallery(label="Generated Image Final")

    time_output = gr.Textbox(label="Processing Time")
    
    generate_btn.click(
        fn=generate_paris_images,
        inputs=[product_name, image1_input, image2_input, woman_prompt, man_prompt, girl_name, girl_hair_length, girl_hair_style, girl_hair_color, boy_name, boy_hair_length, boy_hair_style, boy_hair_color, batch_size],
        outputs=[image_output, image_output_processed, time_output]
    )

    product_name.change(
        fn=change_product_preview,
        inputs=[product_name],
        outputs=[product_preview, woman_prompt, man_prompt]
    )

if __name__ == "__main__":
    print("Starting Gradio interface...")
    demo.launch()