app.py

import gradio as gr
import os
from huggingface_hub import InferenceClient
from pathlib import Path
import tempfile

# Initialize the inference client
client = InferenceClient(
    provider="fal-ai",
    api_key=os.environ.get("HF_TOKEN"),
    bill_to="huggingface",
)

def generate_video_with_auth(image, prompt, profile: gr.OAuthProfile | None, progress=gr.Progress()):
    """
    Generate a video from an image using the Ovi model with authentication check.
    
    Args:
        image: Input image (PIL Image or file path)
        prompt: Text prompt describing the desired motion/animation
        profile: OAuth profile for authentication
        progress: Gradio progress tracker
    
    Returns:
        Path to the generated video file
    """
    if profile is None:
        raise gr.Error("Click Sign in with Hugging Face button to use this app for free")
    
    if image is None:
        raise gr.Error("Please upload an image first!")
    
    if not prompt or prompt.strip() == "":
        raise gr.Error("Please enter a prompt describing the desired motion!")
    
    try:
        progress(0.2, desc="Processing image...")
        
        # Read the image file
        if isinstance(image, str):
            with open(image, "rb") as image_file:
                input_image = image_file.read()
        else:
            # If image is a PIL Image, save it temporarily
            temp_image = tempfile.NamedTemporaryFile(delete=False, suffix=".png")
            image.save(temp_image.name)
            with open(temp_image.name, "rb") as image_file:
                input_image = image_file.read()
        
        progress(0.4, desc="Generating video with AI...")
        
        # Generate video using the inference client
        video = client.image_to_video(
            input_image,
            prompt=prompt,
            model="chetwinlow1/Ovi",
        )
        
        progress(0.9, desc="Finalizing video...")
        
        # Save the video to a temporary file
        output_path = tempfile.NamedTemporaryFile(delete=False, suffix=".mp4")
        
        # Check if video is bytes or a file path
        if isinstance(video, bytes):
            with open(output_path.name, "wb") as f:
                f.write(video)
        elif isinstance(video, str) and os.path.exists(video):
            # If it's a path, copy it
            import shutil
            shutil.copy(video, output_path.name)
        else:
            # Try to write it directly
            with open(output_path.name, "wb") as f:
                f.write(video)
        
        progress(1.0, desc="Complete!")
        
        return output_path.name
    
    except Exception as e:
        raise gr.Error(f"Error generating video: {str(e)}")

# Create the Gradio interface
with gr.Blocks(
    theme=gr.themes.Soft(
        primary_hue="blue",
        secondary_hue="indigo",
    ),
    css="""
        .header-link {
            font-size: 0.9em;
            color: #666;
            text-decoration: none;
            margin-bottom: 1em;
            display: inline-block;
        }
        .header-link:hover {
            color: #333;
            text-decoration: underline;
        }
        .main-header {
            text-align: center;
            margin-bottom: 2em;
        }
        .info-box {
            background-color: #f0f7ff;
            border-left: 4px solid #4285f4;
            padding: 1em;
            margin: 1em 0;
            border-radius: 4px;
        }
        .auth-warning {
            color: #ff6b00;
            font-weight: bold;
            text-align: center;
            margin: 1em 0;
        }
    """,
    title="Image to Video Generator with Ovi",
) as demo:
    
    gr.HTML(
        """
        <div class="main-header">
            <a href="https://huggingface.co/spaces/akhaliq/anycoder" target="_blank" class="header-link">
                Built with anycoder ✨
            </a>
        </div>
        """
    )
    
    gr.Markdown(
        """
        # 🎬 Image to Video Generator with Ovi
        
        Transform your static images into dynamic videos with synchronized audio using AI! Upload an image and describe the motion you want to see.
        
        Powered by **Ovi: Twin Backbone Cross-Modal Fusion for Audio-Video Generation** via HuggingFace Inference API.
        """
    )
    
    gr.HTML(
        """
        <div class="auth-warning">
            ⚠️ You must Sign in with Hugging Face using the button below to use this app.
        </div>
        """
    )
    
    # Add login button - required for OAuth
    gr.LoginButton()
    
    gr.HTML(
        """
        <div class="info-box">
            <strong>💡 Tips for best results:</strong>
            <ul>
                <li>Use clear, well-lit images with a single main subject</li>
                <li>Write specific prompts describing the desired motion or action</li>
                <li>Keep prompts concise and focused on movement and audio elements</li>
                <li>Processing generates 5-second videos at 24 FPS with synchronized audio</li>
                <li>Processing may take 30-60 seconds depending on server load</li>
            </ul>
        </div>
        """
    )
    
    gr.HTML(
        """
        <div class="info-box">
            <strong>✨ Special Tokens for Enhanced Control:</strong>
            <ul>
                <li><strong>Speech:</strong> <code>&lt;S&gt;Your speech content here&lt;E&gt;</code> - Text enclosed in these tags will be converted to speech</li>
                <li><strong>Audio Description:</strong> <code>&lt;AUDCAP&gt;Audio description here&lt;ENDAUDCAP&gt;</code> - Describes the audio or sound effects present in the video</li>
            </ul>
            <br>
            <strong>📝 Example Prompt:</strong><br>
            <code>Dogs bark loudly at a man wearing a red shirt. The man says &lt;S&gt;Please stop barking at me!&lt;E&gt;. &lt;AUDCAP&gt;Dogs barking, angry man yelling in stern voice&lt;ENDAUDCAP&gt;.</code>
        </div>
        """
    )
    
    with gr.Row():
        with gr.Column(scale=1):
            image_input = gr.Image(
                label="📸 Upload Image",
                type="filepath",
                sources=["upload", "clipboard"],
                height=400,
            )
            
            prompt_input = gr.Textbox(
                label="✍️ Text Prompt",
                lines=3,
            )
            
            generate_btn = gr.Button(
                "🎬 Generate Video",
                variant="primary",
                size="lg",
            )
            
            clear_btn = gr.Button(
                "🗑️ Clear",
                variant="secondary",
            )
            
            gr.Examples(
                examples=[
                    [
                        "5.png",
                        'A bearded man wearing large dark sunglasses and a blue patterned cardigan sits in a studio, actively speaking into a large, suspended microphone. He has headphones on and gestures with his hands, displaying rings on his fingers. Behind him, a wall is covered with red, textured sound-dampening foam on the left, and a white banner on the right features the "CHOICE FM" logo and various social media handles like "@ilovechoicefm" with "RALEIGH" below it. The man intently addresses the microphone, articulating, <S>is talent. It\'s all about authenticity. You gotta be who you really are, especially if you\'re working<E>. He leans forward slightly as he speaks, maintaining a serious expression behind his sunglasses.. <AUDCAP>Clear male voice speaking into a microphone, a low background hum.<ENDAUDCAP>'
                    ]
                ],
                inputs=[image_input, prompt_input],
                label="Example",
            )
        
        with gr.Column(scale=1):
            video_output = gr.Video(
                label="🎥 Generated Video",
                height=400,
                autoplay=True,
            )
            
            gr.Markdown(
                """
                ### About Ovi Model
                
                **Ovi: Twin Backbone Cross-Modal Fusion for Audio-Video Generation**
                
                Developed by Chetwin Low, Weimin Wang (Character AI) & Calder Katyal (Yale University)
                
                🌟 **Key Features:**
                - 🎬 **Video+Audio Generation**: Generates synchronized video and audio content simultaneously
                - 📝 **Flexible Input**: Supports text-only or text+image conditioning  
                - ⏱️ **5-second Videos**: Generates 5-second videos at 24 FPS
                - 📐 **Multiple Aspect Ratios**: Supports 720×720 area at various ratios (9:16, 16:9, 1:1, etc)
                
                Ovi is a veo-3 like model that simultaneously generates both video and audio content from text or text+image inputs.
                """
            )
    
    # Event handlers with authentication
    generate_btn.click(
        fn=generate_video_with_auth,
        inputs=[image_input, prompt_input],
        outputs=[video_output],
        queue=False,
        api_name=False,
        show_api=False,
    )
    
    clear_btn.click(
        fn=lambda: (None, "", None),
        inputs=None,
        outputs=[image_input, prompt_input, video_output],
        queue=False,
    )
    
    gr.Markdown(
        """
        ---
        
        ### 🚀 How it works
        
        1. **Sign in** with your Hugging Face account
        2. **Upload** your image - any photo or illustration
        3. **Describe** the motion you want to see in the prompt
        4. **Generate** and watch your image come to life!
        
        ### ⚠️ Notes
        
        - Video generation may take 30-60 seconds
        - Generates 5-second videos at 24 FPS with synchronized audio
        - Supports multiple aspect ratios (9:16, 16:9, 1:1, etc) at 720×720 area
        - Requires a valid HuggingFace token with Inference API access
        - Best results with clear, high-quality images
        - The model works best with realistic subjects and natural motions
        
        ### 🔗 Resources
        
        - [Ovi Model Card](https://huggingface.co/chetwinlow1/Ovi)
        - [HuggingFace Inference API](https://huggingface.co/docs/huggingface_hub/guides/inference)
        - [Character AI](https://character.ai)
        """
    )

# Launch the app
if __name__ == "__main__":
    demo.launch(
        show_api=False,
        enable_monitoring=False,
        quiet=True,
    )