app.py

import gradio as gr
from pytube import YouTube
import subprocess
import torch
from huggingsound import SpeechRecognitionModel
import librosa
import soundfile as sf
from transformers import pipeline

def summarize_video(youtube_link):
    # Download YouTube video's audio
    yt = YouTube(youtube_link)
    yt.streams.filter(only_audio=True, file_extension='mp4').first().download(filename='ytaudio.mp4')
    
    # Convert to WAV format
    subprocess.run(['ffmpeg', '-i', 'ytaudio.mp4', '-acodec', 'pcm_s16le', '-ar', '16000', 'ytaudio.wav'], check=True)
    
    # Initialize speech recognition model
    device = "cuda" if torch.cuda.is_available() else "cpu"
    model = SpeechRecognitionModel("jonatasgrosman/wav2vec2-large-xlsr-53-english", device=device)
    
    # Process audio file and transcribe
    input_file = 'ytaudio.wav'
    stream = librosa.stream(input_file, block_length=30, frame_length=16000, hop_length=16000)
    full_transcript = ''
    for i, speech in enumerate(stream):
        sf.write(f'{i}.wav', speech, 16000)
        transcription = model.transcribe([f'{i}.wav'])
        full_transcript += ' '.join([item['transcription'] for item in transcription])
    
    # Summarize the transcript
    summarizer = pipeline('summarization')
    summarized_text = summarizer(full_transcript, max_length=130, min_length=30, do_sample=False)
    return summarized_text[0]['summary_text']

# Set up the Gradio interface
iface = gr.Interface(fn=summarize_video,
                     inputs=gr.inputs.Textbox(lines=2, placeholder="Enter YouTube Video Link Here..."),
                     outputs="text",
                     title="YouTube Video Text Summarizer",
                     description="This tool summarizes the text extracted from a given YouTube video. Please enter the video link below.")

if __name__ == "__main__":
    iface.launch()