import os import cv2 import numpy as np import matplotlib.pyplot as plt from moviepy.editor import VideoFileClip, AudioFileClip import librosa import librosa.display import soundfile as sf import gradio as gr import tempfile # Function for displaying progress def display_progress(percent, message, progress=gr.Progress()): progress(percent, desc=message) # Function for extracting audio from video def extract_audio(video_path, progress): display_progress(0.1, "Extracting audio from video", progress) video = VideoFileClip(video_path) audio_path = "extracted_audio.wav" video.audio.write_audiofile(audio_path) display_progress(0.2, "Audio extracted", progress) return audio_path # Function for dividing video into frames def extract_frames(video_path, progress): display_progress(0.3, "Extract frames from video", progress) video = cv2.VideoCapture(video_path) frames = [] success, frame = video.read() while success: frames.append(frame) success, frame = video.read() video.release() display_progress(0.4, "Frames extracted", progress) return frames # Convert frame to spectrogram def frame_to_spectrogram(frame, sr=22050): gray_frame = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY) S = np.flipud(gray_frame.astype(np.float32) / 255.0 * 100.0) y = librosa.griffinlim(S) return y # Saving audio def save_audio(y, sr=22050): audio_path = 'output_frame_audio.wav' sf.write(audio_path, y, sr) return audio_path # Saving frame spectrogram def save_spectrogram_image(S, frame_number, temp_dir): plt.figure(figsize=(10, 4)) librosa.display.specshow(S) plt.tight_layout() image_path = os.path.join(temp_dir, f'spectrogram_frame_{frame_number}.png') plt.savefig(image_path) plt.close() return image_path # Processing all video frames def process_video_frames(frames, sr=22050, temp_dir=None, progress=gr.Progress()): processed_frames = [] total_frames = len(frames) for i, frame in enumerate(frames): y = frame_to_spectrogram(frame, sr) S = librosa.feature.melspectrogram(y=y, sr=sr) image_path = save_spectrogram_image(S, i, temp_dir) processed_frame = cv2.imread(image_path) processed_frames.append(processed_frame) display_progress(0.5 + int((i + 1) / total_frames * 0.7), f"Frame processing {i + 1}/{total_frames}", progress) display_progress(0.8, "All frames processed", progress) return processed_frames # Saving video from frames def save_video_from_frames(frames, output_path, fps=30): height, width, layers = frames[0].shape video = cv2.VideoWriter(output_path, cv2.VideoWriter_fourcc(*'mp4v'), fps, (width, height)) for frame in frames: video.write(frame) video.release() # Adding audio back to video def add_audio_to_video(video_path, audio_path, output_path, progress): display_progress(0.9, "Adding audio back to video", progress) video = VideoFileClip(video_path) audio = AudioFileClip(audio_path) final_video = video.set_audio(audio) final_video.write_videofile(output_path, codec='libx264', audio_codec='aac') display_progress(1, "Video's ready", progress) # Gradio interface def process_video(video_path, progress=gr.Progress()): audio_path = extract_audio(video_path, progress) frames = extract_frames(video_path, progress) # Creating a temporary folder for saving frames with tempfile.TemporaryDirectory() as temp_dir: processed_frames = process_video_frames(frames, temp_dir=temp_dir, progress=progress) temp_video_path = os.path.join(temp_dir, 'processed_video.mp4') save_video_from_frames(processed_frames, temp_video_path) output_video_path = 'output_video_with_audio.mp4' add_audio_to_video(temp_video_path, audio_path, output_video_path, progress) return output_video_path with gr.Blocks(title='Video from Spectrogram', theme=gr.themes.Soft(primary_hue="green", secondary_hue="green", spacing_size="sm", radius_size="lg")) as iface: with gr.Group(): with gr.Row(variant='panel'): with gr.Column(): gr.HTML("

Telegram Channel

") with gr.Column(): gr.HTML("

Telegram Chat

") with gr.Column(): gr.HTML("

YouTube

") with gr.Column(): gr.HTML("

GitHub

") with gr.Column(variant='panel'): video_input = gr.Video(label="Upload video") with gr.Column(variant='panel'): generate_button = gr.Button("Generate") with gr.Column(variant='panel'): video_output = gr.Video(label="VideoSpectrogram") def gradio_video_process_fn(video_input, progress=gr.Progress()): return process_video(video_input, progress) generate_button.click( gradio_video_process_fn, inputs=[video_input], outputs=[video_output] ) iface.launch(share=True)