app.py

import gradio as gr
import os
from huggingface_hub import InferenceClient
import tempfile
import shutil
from pathlib import Path
from typing import Optional, Union
import time

# -------------------------
# Utilities
# -------------------------

def cleanup_temp_files():
    try:
        temp_dir = tempfile.gettempdir()
        for file_path in Path(temp_dir).glob("*.mp4"):
            try:
                if file_path.stat().st_mtime < (time.time() - 300):
                    file_path.unlink(missing_ok=True)
            except Exception:
                pass
    except Exception as e:
        print(f"Cleanup error: {e}")

def _client_from_token(token: Optional[str]) -> InferenceClient:
    if not token:
        raise gr.Error("Please sign in first. This app requires your Hugging Face login.")
    # IMPORTANT: do not set bill_to when using user OAuth tokens
    return InferenceClient(
        provider="fal-ai",
        api_key=token,
    )

def _save_bytes_as_temp_mp4(data: bytes) -> str:
    temp_file = tempfile.NamedTemporaryFile(suffix=".mp4", delete=False)
    try:
        temp_file.write(data)
        temp_file.flush()
        return temp_file.name
    finally:
        temp_file.close()

def text_to_video(prompt, token: gr.OAuthToken | None, duration=5, aspect_ratio="16:9", resolution="720p", *_):
    """Generate video from text prompt"""
    try:
        if token is None or not getattr(token, "token", None):
            return None, "❌ Sign in with Hugging Face to continue. This app uses your inference provider credits."
        
        if not prompt or prompt.strip() == "":
            return None, "Please enter a text prompt"
        
        cleanup_temp_files()
        
        # Create client with user's token
        client = _client_from_token(token.token)
        
        # Generate video from text
        try:
            video = client.text_to_video(
                prompt,
                model="akhaliq/veo3.1-fast",
            )
        except Exception as e:
            import requests
            if isinstance(e, requests.HTTPError) and getattr(e.response, "status_code", None) == 403:
                return None, "❌ Access denied by provider (403). Make sure your HF account has credits/permission for provider 'fal-ai' and model 'akhaliq/veo3.1-fast'."
            raise
        
        # Save the video to a temporary file
        video_path = _save_bytes_as_temp_mp4(video)
        
        return video_path, f"✅ Video generated successfully from prompt: '{prompt[:50]}...'"
    
    except gr.Error as e:
        return None, f"❌ {str(e)}"
    except Exception as e:
        return None, f"❌ Generation failed. If this keeps happening, check your provider quota or try again later."

def image_to_video(image, prompt, token: gr.OAuthToken | None, duration=5, aspect_ratio="16:9", resolution="720p", *_):
    """Generate video from image and prompt"""
    try:
        if token is None or not getattr(token, "token", None):
            return None, "❌ Sign in with Hugging Face to continue. This app uses your inference provider credits."
        
        if image is None:
            return None, "Please upload an image"
        
        if not prompt or prompt.strip() == "":
            return None, "Please enter a prompt describing the motion"
        
        cleanup_temp_files()
        
        # Read the image file
        if isinstance(image, str):
            # If image is a file path
            with open(image, "rb") as image_file:
                input_image = image_file.read()
        else:
            # If image is already bytes or similar
            import io
            from PIL import Image as PILImage
            
            # Convert to bytes if necessary
            if isinstance(image, PILImage.Image):
                buffer = io.BytesIO()
                image.save(buffer, format='PNG')
                input_image = buffer.getvalue()
            else:
                # Assume it's a numpy array or similar
                pil_image = PILImage.fromarray(image)
                buffer = io.BytesIO()
                pil_image.save(buffer, format='PNG')
                input_image = buffer.getvalue()
        
        # Create client with user's token
        client = _client_from_token(token.token)
        
        # Generate video from image
        try:
            video = client.image_to_video(
                input_image,
                prompt=prompt,
                model="akhaliq/veo3.1-fast-image-to-video",
            )
        except Exception as e:
            import requests
            if isinstance(e, requests.HTTPError) and getattr(e.response, "status_code", None) == 403:
                return None, "❌ Access denied by provider (403). Make sure your HF account has credits/permission for provider 'fal-ai' and model 'akhaliq/veo3.1-fast-image-to-video'."
            raise
        
        # Save the video to a temporary file
        video_path = _save_bytes_as_temp_mp4(video)
        
        return video_path, f"✅ Video generated successfully with motion: '{prompt[:50]}...'"
    
    except gr.Error as e:
        return None, f"❌ {str(e)}"
    except Exception as e:
        return None, f"❌ Generation failed. If this keeps happening, check your provider quota or try again later."

def clear_text_tab():
    """Clear text-to-video tab"""
    return "", None, ""

def clear_image_tab():
    """Clear image-to-video tab"""
    return None, "", None, ""

# Custom CSS for better styling
custom_css = """
.container {
    max-width: 1200px;
    margin: auto;
}
.header-link {
    text-decoration: none;
    color: #2196F3;
    font-weight: bold;
}
.header-link:hover {
    text-decoration: underline;
}
.status-box {
    padding: 10px;
    border-radius: 5px;
    margin-top: 10px;
}
.notice {
    background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
    color: white;
    padding: 14px 16px;
    border-radius: 12px;
    margin: 18px auto 6px;
    max-width: 860px;
    text-align: center;
    font-size: 0.98rem;
}
.mobile-link-container {
    background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
    padding: 1.5em;
    border-radius: 10px;
    text-align: center;
    margin: 1em 0;
    box-shadow: 0 4px 6px rgba(0, 0, 0, 0.1);
}
.mobile-link {
    color: white !important;
    font-size: 1.2em;
    font-weight: bold;
    text-decoration: none;
    display: inline-block;
    padding: 0.5em 1.5em;
    background: rgba(255, 255, 255, 0.2);
    border-radius: 25px;
    transition: all 0.3s ease;
}
.mobile-link:hover {
    background: rgba(255, 255, 255, 0.3);
    transform: translateY(-2px);
    box-shadow: 0 4px 8px rgba(0, 0, 0, 0.2);
}
.mobile-text {
    color: white;
    margin-bottom: 0.5em;
    font-size: 1.1em;
}
"""

# Create the Gradio interface
with gr.Blocks(css=custom_css, theme=gr.themes.Soft(), title="AI Video Generator (Paid)") as demo:
    gr.Markdown(
        """
        # 🎬 AI Video Generator
        ### Generate stunning videos from text or animate your images with AI
        #### Powered by VEO 3.1 Fast Model via Hugging Face Inference API (provider: fal-ai)
        """
    )
    
    gr.HTML(
        """
        <div style="text-align:center; max-width:900px; margin:0 auto;">
            <h1 style="font-size:2.2em; margin-bottom:6px;">🎬 Sora-2</h1>
            <p style="color:#777; margin:0 0 8px;">Generate videos via the Hugging Face Inference API (provider: fal-ai)</p>
            <div class="notice">
                <b>Heads up:</b> This is a paid app that uses <b>your</b> inference provider credits when you run generations.
                Free users get <b>$0.10 in included credits</b>. <b>PRO users</b> get <b>$2 in included credits</b> 
                and can continue using beyond that (with billing). 
                <a href='http://huggingface.co/subscribe/pro?source=veo3' target='_blank' style='color:#fff; text-decoration:underline; font-weight:bold;'>Subscribe to PRO</a> 
                for more credits. Please sign in with your Hugging Face account to continue.
            </div>
            <p style="font-size: 0.9em; color: #999; margin-top: 10px;">
                Built with <a href="https://huggingface.co/spaces/akhaliq/anycoder" target="_blank" style="color:#fff; text-decoration:underline;">anycoder</a>
            </p>
        </div>
        """
    )
    
    # Add mobile link section
    gr.HTML(
        """
        <div class="mobile-link-container">
            <div class="mobile-text">📱 On mobile? Use the optimized version:</div>
            <a href="https://akhaliq-veo3-1-fast.hf.space" target="_blank" class="mobile-link">
                🚀 Open Mobile Version
            </a>
        </div>
        """
    )
    
    gr.HTML(
        """
        <p style="text-align: center; font-size: 0.9em; color: #999; margin-top: 10px;">
            Built with <a href="https://huggingface.co/spaces/akhaliq/anycoder" target="_blank" style="color:#667eea; text-decoration:underline;">anycoder</a>
        </p>
        """
    )
    
    # Add login button - required for OAuth
    login_btn = gr.LoginButton("Sign in with Hugging Face")
    
    with gr.Tabs() as tabs:
        # Text-to-Video Tab
        with gr.Tab("📝 Text to Video", id=0):
            gr.Markdown("### Transform your text descriptions into dynamic videos")
            
            with gr.Row():
                with gr.Column(scale=1):
                    text_prompt = gr.Textbox(
                        label="Text Prompt",
                        placeholder="Describe the video you want to create... (e.g., 'A young man walking on the street during sunset')",
                        lines=4,
                        max_lines=6
                    )
                    
                    with gr.Row():
                        text_generate_btn = gr.Button("🎬 Generate Video", variant="primary", scale=2)
                        text_clear_btn = gr.ClearButton(value="🗑️ Clear", scale=1)
                    
                    text_status = gr.Textbox(
                        label="Status",
                        interactive=False,
                        visible=True,
                        elem_classes=["status-box"]
                    )
                
                with gr.Column(scale=1):
                    text_video_output = gr.Video(
                        label="Generated Video",
                        autoplay=True,
                        show_download_button=True,
                        height=400
                    )
            
            # Examples for text-to-video
            gr.Examples(
                examples=[
                    ["A serene beach at sunset with gentle waves"],
                    ["A bustling city street with neon lights at night"],
                    ["A majestic eagle soaring through mountain peaks"],
                    ["An astronaut floating in space near the International Space Station"],
                    ["Cherry blossoms falling in slow motion in a Japanese garden"],
                ],
                inputs=text_prompt,
                label="Example Prompts"
            )
        
        # Image-to-Video Tab
        with gr.Tab("🖼️ Image to Video", id=1):
            gr.Markdown("### Bring your static images to life with motion")
            
            with gr.Row():
                with gr.Column(scale=1):
                    image_input = gr.Image(
                        label="Upload Image",
                        type="pil",
                        height=300
                    )
                    
                    image_prompt = gr.Textbox(
                        label="Motion Prompt",
                        placeholder="Describe how the image should move... (e.g., 'The cat starts to dance')",
                        lines=3,
                        max_lines=5
                    )
                    
                    with gr.Row():
                        image_generate_btn = gr.Button("🎬 Animate Image", variant="primary", scale=2)
                        image_clear_btn = gr.ClearButton(value="🗑️ Clear", scale=1)
                    
                    image_status = gr.Textbox(
                        label="Status",
                        interactive=False,
                        visible=True,
                        elem_classes=["status-box"]
                    )
                
                with gr.Column(scale=1):
                    image_video_output = gr.Video(
                        label="Generated Video",
                        autoplay=True,
                        show_download_button=True,
                        height=400
                    )
            
            # Examples for image-to-video
            gr.Examples(
                examples=[
                    [None, "The person starts walking forward"],
                    [None, "The animal begins to run"],
                    [None, "Camera slowly zooms in while the subject smiles"],
                    [None, "The flowers sway gently in the breeze"],
                    [None, "The clouds move across the sky in time-lapse"],
                ],
                inputs=[image_input, image_prompt],
                label="Example Motion Prompts"
            )
    
    # How to Use section
    with gr.Accordion("📖 How to Use", open=False):
        gr.Markdown(
            """
            ### Text to Video:
            1. Enter a detailed description of the video you want to create
            2. Optionally adjust advanced settings (duration, aspect ratio, resolution)
            3. Click "Generate Video" and wait for the AI to create your video
            4. Download or preview your generated video
            
            ### Image to Video:
            1. Upload an image you want to animate
            2. Describe the motion or action you want to add to the image
            3. Optionally adjust advanced settings
            4. Click "Animate Image" to bring your image to life
            5. Download or preview your animated video
            
            ### Tips for Better Results:
            - Be specific and descriptive in your prompts
            - For image-to-video, describe natural motions that fit the image
            - Use high-quality input images for better results
            - Experiment with different prompts to get the desired effect
            
            ### Mobile Users:
            - For the best mobile experience, use the optimized version at: https://akhaliq-veo3-1-fast.hf.space
            """
        )
    
    # Event handlers - FIXED: removed login_btn from inputs
    text_generate_btn.click(
        fn=text_to_video,
        inputs=[text_prompt],
        outputs=[text_video_output, text_status],
        show_progress="full",
        queue=False,
        api_name=False,
        show_api=False
    )
    
    text_clear_btn.click(
        fn=clear_text_tab,
        inputs=[],
        outputs=[text_prompt, text_video_output, text_status],
        queue=False
    )
    
    image_generate_btn.click(
        fn=image_to_video,
        inputs=[image_input, image_prompt],
        outputs=[image_video_output, image_status],
        show_progress="full",
        queue=False,
        api_name=False,
        show_api=False
    )
    
    image_clear_btn.click(
        fn=clear_image_tab,
        inputs=[],
        outputs=[image_input, image_prompt, image_video_output, image_status],
        queue=False
    )

# Launch the app
if __name__ == "__main__":
    try:
        cleanup_temp_files()
        if os.path.exists("gradio_cached_examples"):
            shutil.rmtree("gradio_cached_examples", ignore_errors=True)
    except Exception as e:
        print(f"Initial cleanup error: {e}")
    
    demo.queue(status_update_rate="auto", api_open=False, default_concurrency_limit=None)
    demo.launch(
        show_api=False,
        share=False,
        show_error=True,
        enable_monitoring=False,
        quiet=True,
        ssr_mode=True
    )