app.py

import gradio as gr
import spaces
import torch
import json
from pathlib import Path
from PIL import Image
import numpy as np

def ensure_env_installed():
    try:
        import transformers
        import torchvision
        import diffusers
        import einops
    except ImportError:
        import subprocess
        import sys
        subprocess.check_call([sys.executable, "-m", "pip", "install", 
                               "transformers==4.54.0",
                               "torchvision==0.22.1",
                               "diffusers==0.34.0",
                               "einops==0.8.1"])

ensure_env_installed()

# Global model variable
model_zoo = {
    "imuru_small": {
        "repo_id": "Ruian7P/imuru_small",
    },
    "imuru_large": {
        "repo_id": "Ruian7P/imuru_large",
    },
    # "emuru_t5_small": {
    #     "repo_id": "Ruian7P/emuru_result",
    #     "model_name": "emuru_t5_small_2e-5_ech5"
    # }
}

model = None

def load_model(model_name="imuru_large"):    
    global model

    if model is None:
        print(f"Loading model {model_name}...")
        from transformers import AutoModel
        
        model = AutoModel.from_pretrained(
            model_zoo[model_name]["repo_id"],
            trust_remote_code=True
        )
        model.eval()
        print("✅ Model loaded")
    
    return model


def load_examples():
    """Load example samples."""
    examples = []
    examples.append([
        "sample/sample.png", "Ruian7P"
    ])
    return examples

def process_image(img):
    from torchvision.transforms import functional as F
    img = img.convert("RGB")
    img = img.resize((img.width * 64 // img.height, 64))
    img = F.to_tensor(img)
    img = F.normalize(img, [0.5], [0.5])
    return img


@spaces.GPU
def generate_handwriting(style_image, gen_text, model_name="imuru_large"):
    """Generate handwriting in the style of the input image."""
    if not gen_text or gen_text.strip() == "":
        return None, "❌ Please provide text to generate"
    
    if style_image is None:
        return None, "❌ Please upload a style image"
    
    try:
        # Convert numpy array to PIL Image if needed
        if isinstance(style_image, np.ndarray):
            style_image = Image.fromarray(style_image)
        
        # Load and move model to GPU
        loaded_model = load_model(model_name)
        loaded_model.to("cuda")
        
        # Preprocess style image
        style_img = process_image(style_image).to("cuda")
        
        # Generate
        with torch.inference_mode():
            result = loaded_model.generate(
                style_img=style_img,
                gen_text=gen_text,
                max_new_tokens=512
            )
        
        return result, "✅ Generation successful!"
        
    except Exception as e:
        import traceback
        traceback.print_exc()
        return None, f"❌ Error: {str(e)}"


# Custom CSS for better styling
custom_css = """
.gradio-container {
    width: 100%;
    max-width: 1200px !important;
    margin: 0 auto !important;
}
.header-text {
    text-align: center;
    margin-bottom: 1rem;
}
.feature-box {
    background: linear-gradient(135deg, #1a1a2e 0%, #16213e 100%);
    border-radius: 10px;
    padding: 15px;
    margin: 10px 0;
}
footer {
    visibility: hidden;
}
"""

# Build the interface with gr.Blocks for better customization
with gr.Blocks(css=custom_css, title="Imuru") as demo:
    
    # Header
    gr.HTML("""
    <div style="text-align: center; margin-bottom: 20px;">
        <h1>🍎 Imuru: Autoregressive Handwriting Generation</h1>
    </div>
    """)
    
    with gr.Row():
        with gr.Column(scale=1):
            model_selector = gr.Dropdown(
                label="🤖 Select Model",
                choices=list(model_zoo.keys()),
                value="imuru_large",
                interactive=True
            )

            style_image_input = gr.Image(
                label="🖼️ Style Image",
                type="pil",
                height=200
            )
            
            gen_text_input = gr.Textbox(
                label="✍️ Text to Generate",
                placeholder="Enter the text you want to generate in the selected style",
                lines=2,
                value="Hello, I am Imuru!"
            )
            
            generate_btn = gr.Button("🕶 Generate", variant="primary", size="lg")
        
        with gr.Column(scale=1):
            output_image = gr.Image(
                label="🖼️ Generated Output",
                type="pil",
                height=200
            )
            
            status_text = gr.Textbox(
                label="🚧 Status",
                lines=1,
                interactive=False
            )
    
    # Examples
    examples = load_examples()
    if examples:
        gr.Examples(
            examples=examples,
            inputs=[style_image_input, gen_text_input],
            label="💡 Examples",
            examples_per_page=4
        )
    
    # Connect events
    generate_btn.click(
        fn=generate_handwriting,
        inputs=[style_image_input, gen_text_input, model_selector],
        outputs=[output_image, status_text]
    )
    
    gen_text_input.submit(
        fn=generate_handwriting,
        inputs=[style_image_input, gen_text_input, model_selector],
        outputs=[output_image, status_text]
    )
    
    # How to use section
    gr.Markdown("""
    ---
    ### 🧠 How to Use
    
    1. **Upload a style image**: A handwritten sample to extract style from
    2. **Type generation text**: The text you want to generate in the style of the image
    3. **Click Generate**: Imuru will create the handwritten text image for you!
    """)


if __name__ == "__main__":
    demo.launch()