app_exp2.py

import gradio as gr
import vtracer
import os
import pandas as pd
from io import BytesIO
from PIL import Image
import cairosvg
import cv2
import numpy as np
import tempfile

def clean_svg(svg_string):
    """Optional function to clean SVG if needed"""
    return svg_string

def rasterize_svg(svg_string, width, height, dpi=128, scale=1):
    """Convert SVG string to PNG image while maintaining aspect ratio"""
    try:
        svg_raster_bytes = cairosvg.svg2png(
            bytestring=svg_string,
            background_color='white',
            output_width=width,
            output_height=height,
            dpi=dpi,
            scale=scale)
        svg_raster = Image.open(BytesIO(svg_raster_bytes))
    except:
        try:
            svg = clean_svg(svg_string)
            svg_raster_bytes = cairosvg.svg2png(
                bytestring=svg,
                background_color='white',
                output_width=width,
                output_height=height,
                dpi=dpi,
                scale=scale)
            svg_raster = Image.open(BytesIO(svg_raster_bytes))
        except:
            svg_raster = Image.new('RGB', (width, height), color='white')
    return svg_raster

def create_video_from_frames(frame_files, output_path, duration_seconds, width, height):
    """Create video from sequence of frames with specified duration"""
    # Calculate frame rate based on desired duration
    num_frames = len(frame_files)
    fps = max(1, num_frames / duration_seconds)  # Ensure at least 1 fps

    # Initialize video writer
    fourcc = cv2.VideoWriter_fourcc(*'mp4v')
    video = cv2.VideoWriter(output_path, fourcc, fps, (width, height))

    # Read each frame and write to video
    for frame_file in frame_files:
        # Read image with PIL and convert to OpenCV format
        pil_img = Image.open(frame_file)
        cv_img = cv2.cvtColor(np.array(pil_img), cv2.COLOR_RGB2BGR)
        video.write(cv_img)

    # Add last frame to fill remaining time if needed
    if num_frames > 0:
        remaining_frames = max(0, int(fps * duration_seconds) - num_frames)
        for _ in range(remaining_frames):
            video.write(cv_img)

    video.release()

def process_svg_to_video(input_svg_path, original_width, original_height, video_duration_seconds=10, chunk_size=30):
    """Process SVG file and create a video with specified duration using exponential row slicing"""
    # Read SVG file as a table to maintain exact row slicing logic
    df = pd.read_table(input_svg_path, header=None)
    df_head = df.head(3)
    df_tail = df.tail(1)
    df_middle = df.iloc[3:-1, :]
    
    # Use the original image dimensions
    width, height = original_width, original_height
    
    # If chunk_size is 0, use automatic calculation (start with 1)
    total_rows = len(df_middle)
    if chunk_size == 0:
        initial_chunk = 1  # Start with 1 path element
    else:
        initial_chunk = max(1, min(chunk_size, total_rows))  # Ensure it's within valid range
    
    # Create a temporary directory for images
    temp_dir = tempfile.mkdtemp()
    frame_files = []
    
    # Process with exponential chunk sizes
    current_chunk_size = initial_chunk
    processed_rows = 0
    
    while processed_rows < total_rows:
        # Calculate end index for this chunk
        end_idx = min(processed_rows + current_chunk_size, total_rows)
        current_chunk = df_middle.iloc[:end_idx]
        
        # Combine with head and tail
        combined_df = pd.concat([df_head, current_chunk, df_tail])
        svg_content = "\n".join(combined_df[0].astype(str).values.tolist())
        
        # Convert to image using original dimensions
        img = rasterize_svg(svg_content, width, height)
        img_filename = os.path.join(temp_dir, f"frame_{processed_rows:04d}.png")
        img.save(img_filename)
        frame_files.append(img_filename)
        
        # Update counters
        processed_rows = end_idx
        current_chunk_size *= 2  # Double the chunk size for next iteration
    
    # Create output video path
    output_video_path = os.path.join(temp_dir, "output_video.mp4")
    
    # Create video from frames
    create_video_from_frames(frame_files, output_video_path, video_duration_seconds, width, height)
    
    # Clean up temporary files (except the video)
    for file in frame_files:
        os.remove(file)
    
    return output_video_path, temp_dir

def convert_to_vector_and_video(
    image, 
    video_duration=10,
    chunk_size=30,
    colormode="color", 
    hierarchical="stacked", 
    mode="spline", 
    filter_speckle=4, 
    color_precision=6, 
    layer_difference=16, 
    corner_threshold=60, 
    length_threshold=4.0, 
    max_iterations=10, 
    splice_threshold=45, 
    path_precision=3
):
    # Create temporary directory
    temp_dir = tempfile.mkdtemp()
    input_path = os.path.join(temp_dir, "temp_input.jpg")
    output_svg_path = os.path.join(temp_dir, "svg_output.svg")
    
    # Save the input image to a temporary file
    image.save(input_path)
    
    # Get original dimensions from the uploaded image
    original_width, original_height = image.size
    
    # Convert the image to SVG using VTracer
    vtracer.convert_image_to_svg_py(
        input_path,
        output_svg_path,
        colormode=colormode,
        hierarchical=hierarchical,
        mode=mode,
        filter_speckle=int(filter_speckle),
        color_precision=int(color_precision),
        layer_difference=int(layer_difference),
        corner_threshold=int(corner_threshold),
        length_threshold=float(length_threshold),
        max_iterations=int(max_iterations),
        splice_threshold=int(splice_threshold),
        path_precision=int(path_precision)
    )
    
    # Process SVG to create video using the original dimensions
    video_path, video_temp_dir = process_svg_to_video(
        output_svg_path,
        original_width,
        original_height,
        video_duration_seconds=video_duration,
        chunk_size=chunk_size
    )
    
    # Read the SVG output
    with open(output_svg_path, "r") as f:
        svg_content = f.read()
    
    # Return the SVG preview, SVG file, and video file
    return (
        gr.HTML(f'<svg viewBox="0 0 {original_width} {original_height}">{svg_content}</svg>'),
        output_svg_path,
        video_path
    )

def handle_color_mode(value):
    return value

def clear_inputs():
    return (
        gr.Image(value=None), 
        gr.Slider(value=10),
        gr.Slider(value=30),
        gr.Radio(value="color"), 
        gr.Radio(value="stacked"), 
        gr.Radio(value="spline"), 
        gr.Slider(value=4), 
        gr.Slider(value=6), 
        gr.Slider(value=16), 
        gr.Slider(value=60), 
        gr.Slider(value=4.0), 
        gr.Slider(value=10), 
        gr.Slider(value=45), 
        gr.Slider(value=3)
    )

def update_interactivity_and_visibility(colormode, color_precision_value, layer_difference_value):
    is_color_mode = colormode == "color"
    return (
        gr.update(interactive=is_color_mode),
        gr.update(interactive=is_color_mode),
        gr.update(visible=is_color_mode)
    )

def update_interactivity_and_visibility_for_mode(mode):
    is_spline_mode = mode == "spline"
    return (
        gr.update(interactive=is_spline_mode),
        gr.update(interactive=is_spline_mode),
        gr.update(interactive=is_spline_mode)
    )

css = """
#col-container {
    margin: 0 auto;
    max-width: 960px;
}
.generate-btn {
    background: linear-gradient(90deg, #4B79A1 0%, #283E51 100%) !important;
    border: none !important;
    color: white !important;
}
.generate-btn:hover {
    transform: translateY(-2px);
    box-shadow: 0 5px 15px rgba(0,0,0,0.2);
}
"""

examples = [
    "examples/玉子.jpg",
    "examples/异闻录.jpg",
    "examples/化物语封面.jpeg",
    "examples/01.jpg",
    "examples/02.jpg",
    "examples/03.jpg",
]

# Define the Gradio interface
with gr.Blocks(css=css) as app:
    with gr.Column(elem_id="col-container"):
        gr.HTML("""
            <div style="text-align: center;">
                <h2>Image to Vector Video Converter ⚡</h2>
                <p>Converts raster images to vector graphics and creates progressive rendering videos.</p>
            </div>
        """)
        with gr.Row():
            with gr.Column():
                image_input = gr.Image(type="pil", label="Upload Image")
                video_duration = gr.Slider(1, 60, value=10, step=1, label="Video Duration (seconds)")
                chunk_size = gr.Slider(0, 1000, value=300, step=1, label="Chunk Size (0=auto)", 
                                     info="Number of SVG path elements to add per frame (0 for automatic calculation)")
                
                with gr.Accordion("Advanced Settings", open=False):
                    with gr.Accordion("Clustering", open=False):
                        colormode = gr.Radio([("COLOR","color"),("B/W", "binary")], value="color", label="Color Mode", show_label=False)
                        filter_speckle = gr.Slider(0, 128, value=4, step=1, label="Filter Speckle", info="Cleaner")
                        color_precision = gr.Slider(1, 8, value=6, step=1, label="Color Precision", info="More accurate")
                        layer_difference = gr.Slider(0, 128, value=16, step=1, label="Gradient Step", info="Less layers")
                        hierarchical = gr.Radio([("STACKED","stacked"), ("CUTOUT","cutout")], value="stacked", label="Hierarchical Mode",show_label=False)
                    with gr.Accordion("Curve Fitting", open=False):
                        mode = gr.Radio([("SPLINE","spline"),("POLYGON", "polygon"), ("PIXEL","none")], value="spline", label="Mode", show_label=False)
                        corner_threshold = gr.Slider(0, 180, value=60, step=1, label="Corner Threshold", info="Smoother")
                        length_threshold = gr.Slider(3.5, 10, value=4.0, step=0.1, label="Segment Length", info ="More coarse")
                        splice_threshold = gr.Slider(0, 180, value=45, step=1, label="Splice Threshold", info="Less accurate")
                        max_iterations = gr.Slider(1, 20, value=10, step=1, label="Max Iterations", visible=False)
                        path_precision = gr.Slider(1, 10, value=3, step=1, label="Path Precision", visible=False)
                        output_text = gr.Textbox(label="Selected Mode", visible=False)
                
                with gr.Row():
                    clear_button = gr.Button("Clear")
                    convert_button = gr.Button("✨ Convert to Video", variant='primary', elem_classes=["generate-btn"])

            with gr.Column():
                html = gr.HTML(label="SVG Preview")
                svg_output = gr.File(label="Download SVG")
                video_output = gr.Video(label="Rendering Video")
        
        gr.Examples(
            examples=examples,
            fn=convert_to_vector_and_video,
            inputs=[image_input],
            outputs=[html, svg_output, video_output],
            cache_examples=False,
            run_on_click=True
        )
    
    # Event handlers
    colormode.change(handle_color_mode, inputs=colormode, outputs=output_text)
    hierarchical.change(handle_color_mode, inputs=hierarchical, outputs=output_text)
    mode.change(handle_color_mode, inputs=mode, outputs=output_text)
    
    colormode.change(
        update_interactivity_and_visibility,
        inputs=[colormode, color_precision, layer_difference],
        outputs=[color_precision, layer_difference, hierarchical]
    )

    mode.change(
        update_interactivity_and_visibility_for_mode,
        inputs=[mode],
        outputs=[corner_threshold, length_threshold, splice_threshold]
    )

    clear_button.click(
        clear_inputs,
        outputs=[
            image_input,
            video_duration,
            chunk_size,
            colormode,
            hierarchical,
            mode,
            filter_speckle,
            color_precision,
            layer_difference,
            corner_threshold,
            length_threshold,
            max_iterations,
            splice_threshold,
            path_precision
        ]
    )

    convert_button.click(
        convert_to_vector_and_video,
        inputs=[
            image_input,
            video_duration,
            chunk_size,
            colormode,
            hierarchical,
            mode,
            filter_speckle,
            color_precision,
            layer_difference,
            corner_threshold,
            length_threshold,
            max_iterations,
            splice_threshold,
            path_precision
        ],
        outputs=[html, svg_output, video_output]
    )

# Launch the app
if __name__ == "__main__":
    app.launch(share=True)