app.py

import gradio as gr
import replicate
import os
from PIL import Image
import requests
from io import BytesIO
import time
import tempfile
import base64

# Set up Replicate API key from environment variable
os.environ['REPLICATE_API_TOKEN'] = os.getenv('REPLICATE_API_TOKEN')

def upload_to_imgur(image):
    """
    Upload image to Imgur and return URL
    Alternative: You can use other services like Cloudinary, imgbb, etc.
    """
    import base64
    import json
    
    # Convert PIL image to base64
    buffered = BytesIO()
    image.save(buffered, format="PNG")
    img_base64 = base64.b64encode(buffered.getvalue()).decode()
    
    # Imgur API (anonymous upload)
    headers = {
        'Authorization': 'Client-ID 0d90e8a3e7d8b4e'  # Public client ID for anonymous uploads
    }
    
    response = requests.post(
        'https://api.imgur.com/3/image',
        headers=headers,
        data={'image': img_base64}
    )
    
    if response.status_code == 200:
        data = response.json()
        return data['data']['link']
    else:
        raise Exception(f"Failed to upload to Imgur: {response.status_code}")

def process_images(prompt, image1, image2=None):
    """
    Process uploaded images with Replicate API
    """
    if not image1:
        return None, "Please upload at least one image"
    
    # Check if API token is set
    if not os.getenv('REPLICATE_API_TOKEN'):
        return None, "⚠️ Please set REPLICATE_API_TOKEN environment variable"
    
    try:
        status_message = "📤 Uploading images..."
        
        # Upload images to get public URLs
        image_urls = []
        
        try:
            # Try to upload to Imgur (or your preferred service)
            url1 = upload_to_imgur(image1)
            image_urls.append(url1)
            
            if image2:
                url2 = upload_to_imgur(image2)
                image_urls.append(url2)
                
        except Exception as upload_error:
            # Fallback: Convert to base64 data URIs
            buffered1 = BytesIO()
            image1.save(buffered1, format="PNG")
            img_base64_1 = base64.b64encode(buffered1.getvalue()).decode()
            image_urls.append(f"data:image/png;base64,{img_base64_1}")
            
            if image2:
                buffered2 = BytesIO()
                image2.save(buffered2, format="PNG")
                img_base64_2 = base64.b64encode(buffered2.getvalue()).decode()
                image_urls.append(f"data:image/png;base64,{img_base64_2}")
        
        status_message = "🎨 Processing with nano-banana model..."
        
        # Prepare input matching the exact format from your example
        input_data = {
            "prompt": prompt,
            "image_input": image_urls
        }
        
        # Run the model
        output = replicate.run(
            "google/nano-banana",
            input=input_data
        )
        
        # Handle various output formats
        output_url = None
        
        # Check different possible output formats
        if hasattr(output, 'url'):
            output_url = output.url()
        elif isinstance(output, str):
            output_url = output
        elif isinstance(output, list) and len(output) > 0:
            output_url = output[0]
        elif hasattr(output, '__iter__'):
            try:
                for item in output:
                    if isinstance(item, str) and item.startswith('http'):
                        output_url = item
                        break
            except:
                pass
        
        if not output_url:
            return None, f"❌ Error: No valid output URL found. Response type: {type(output)}"
        
        # Download the generated image
        if hasattr(output, 'read'):
            # If output has a read method, use it
            img_data = output.read()
            img = Image.open(BytesIO(img_data))
        else:
            # Otherwise, download from URL
            response = requests.get(output_url)
            if response.status_code == 200:
                img = Image.open(BytesIO(response.content))
            else:
                return None, f"❌ Error: Failed to download image (Status: {response.status_code})"
        
        return img, f"✅ Image generated successfully! Output URL: {output_url[:50]}..."
        
    except replicate.exceptions.ModelError as e:
        return None, f"❌ Model Error: {str(e)}\n\nMake sure 'google/nano-banana' exists and is accessible."
    except Exception as e:
        error_msg = str(e)
        if "not found" in error_msg.lower():
            return None, "❌ Model 'google/nano-banana' not found. Please check:\n1. Model name is correct\n2. Model is public or you have access\n3. Try format: 'owner/model-name'"
        elif "authentication" in error_msg.lower():
            return None, "❌ Authentication failed. Please check your REPLICATE_API_TOKEN."
        else:
            return None, f"❌ Error: {error_msg}"

# Create Gradio interface with gradient theme
css = """
.gradio-container {
    background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
    font-family: 'Inter', sans-serif;
}
.gr-button {
    background: linear-gradient(135deg, #f093fb 0%, #f5576c 100%);
    border: none;
    color: white;
    font-weight: bold;
    transition: transform 0.2s;
}
.gr-button:hover {
    transform: scale(1.05);
    box-shadow: 0 10px 20px rgba(0,0,0,0.2);
}
.gr-input {
    border-radius: 10px;
    border: 2px solid rgba(255,255,255,0.3);
    background: rgba(255,255,255,0.9);
}
.header-text {
    background: linear-gradient(135deg, #f093fb 0%, #f5576c 100%);
    -webkit-background-clip: text;
    -webkit-text-fill-color: transparent;
    background-clip: text;
    font-size: 2.5em;
    font-weight: bold;
    text-align: center;
    margin-bottom: 20px;
}
.description-text {
    color: white;
    text-align: center;
    font-size: 1.1em;
    margin-bottom: 30px;
    text-shadow: 2px 2px 4px rgba(0,0,0,0.2);
}
"""

# Build the Gradio interface
with gr.Blocks(css=css, theme=gr.themes.Soft()) as demo:
    gr.HTML("""
        <div class="header-text">🎨 AI Image Style Transfer Studio</div>
        <div class="description-text">
            Upload 1-2 images and describe how you want them styled.
            The AI will create a beautiful transformation!
        </div>
    """)
    
    with gr.Row():
        with gr.Column(scale=1):
            gr.Markdown("### 📤 Input Section")
            
            prompt = gr.Textbox(
                label="✏️ Style Prompt",
                placeholder="Describe how you want to style your images...",
                lines=3,
                value="Make the sheets in the style of the logo. Make the scene natural."
            )
            
            with gr.Row():
                image1 = gr.Image(
                    label="Image 1 (Required)",
                    type="pil",
                    height=200
                )
                image2 = gr.Image(
                    label="Image 2 (Optional)",
                    type="pil",
                    height=200
                )
            
            generate_btn = gr.Button(
                "🚀 Generate Styled Image",
                variant="primary",
                size="lg"
            )
            
            gr.Markdown("""
            #### 💡 Tips:
            - Upload high-quality images for best results
            - Be specific in your style description
            - Experiment with different prompts!
            """)
        
        with gr.Column(scale=1):
            gr.Markdown("### 🎯 Output Section")
            
            output_image = gr.Image(
                label="Generated Image",
                type="pil",
                height=400
            )
            
            status = gr.Textbox(
                label="Status",
                interactive=False,
                lines=2
            )
    
    # Examples section
    with gr.Row():
        gr.Examples(
            examples=[
                ["Transform into watercolor painting style", None, None],
                ["Make it look like a vintage photograph", None, None],
                ["Apply cyberpunk neon style", None, None],
                ["Convert to minimalist line art", None, None],
            ],
            inputs=[prompt, image1, image2],
            label="Example Prompts"
        )
    
    # Event handlers
    generate_btn.click(
        fn=process_images,
        inputs=[prompt, image1, image2],
        outputs=[output_image, status],
        api_name="generate"
    )
    
    # Additional information
    gr.Markdown("""
    ---
    ### ⚙️ Setup Instructions:
    
    1. **Set Environment Variable:**
       ```bash
       export REPLICATE_API_TOKEN="your_api_token_here"
       ```
       Get your token from: https://replicate.com/account/api-tokens
    
    2. **Install Required Packages:**
       ```bash
       pip install gradio replicate pillow requests
       ```
    
    3. **Image Hosting Options:**
       - **Option A**: Uses Imgur for free image hosting (default)
       - **Option B**: Falls back to base64 data URIs if upload fails
       - **Option C**: Use your own image hosting service (Cloudinary, S3, etc.)
    
    4. **Model Notes:**
       - Using: `google/nano-banana` model
       - If this model doesn't exist, try:
         - `stability-ai/stable-diffusion`
         - `pharmapsychotic/clip-interrogator`
         - Check available models at: https://replicate.com/explore
    
    ### 🔒 Security:
    - API keys are managed through environment variables
    - Never commit API keys to version control
    - Consider implementing user authentication for production
    """)

# Launch the app
if __name__ == "__main__":
    demo.launch(
        share=True,
        server_name="0.0.0.0",
        server_port=7860,
        show_error=True
    )