app.py

import os
import time
import yt_dlp
import requests

# You can replace these URLs with Hugging Face API URLs
BASE_URL = "https://api-inference.huggingface.co/models/"
HEADERS = {"Authorization": "Bearer YOUR_HUGGINGFACE_API_KEY"}  # Replace with your actual Hugging Face API key

# ✅ Download YouTube Audio
def download_audio(youtube_url, output_file="audio.mp3"):
    ydl_opts = {
        'format': 'bestaudio/best',
        'outtmpl': output_file,
        'postprocessors': [{
            'key': 'FFmpegExtractAudio',
            'preferredcodec': 'mp3',
            'preferredquality': '192',
        }],
        'quiet': False,  # Set to False for more verbose output to help debug
    }

    try:
        with yt_dlp.YoutubeDL(ydl_opts) as ydl:
            print("Downloading audio...")
            ydl.download([youtube_url])
        print(f"Audio downloaded successfully: {output_file}")
        return output_file
    except Exception as e:
        print("Error downloading audio:", str(e))
        return None

# ✅ Upload Audio to Hugging Face STT Model for Transcription
def get_transcription(file_path):
    with open(file_path, "rb") as audio_file:
        audio = audio_file.read()

    data = {"inputs": audio}

    try:
        response = requests.post(f"{BASE_URL}/whisper-large", headers=HEADERS, files={"file": audio})
        if response.status_code == 200:
            return response.json()["text"]
        else:
            print(f"Error during transcription request: {response.status_code}")
            return None
    except Exception as e:
        print(f"Error during transcription request: {str(e)}")
        return None

# ✅ Summarize Transcript using Hugging Face GPT-based Model
def summarize_text(transcript_text):
    data = {
        "inputs": f"Summarize the following text:\n\n{transcript_text}",
    }
    try:
        response = requests.post(f"{BASE_URL}/gpt2", headers=HEADERS, json=data)
        if response.status_code == 200:
            return response.json()[0]['generated_text']
        else:
            print(f"Error summarizing transcript: {response.status_code}")
            return None
    except Exception as e:
        print(f"Error summarizing transcript: {str(e)}")
        return None

# ✅ Text-to-Speech (TTS) using Hugging Face
def generate_tts_audio(summary_text):
    data = {"inputs": summary_text}
    try:
        response = requests.post(f"{BASE_URL}/tacotron2", headers=HEADERS, json=data)
        if response.status_code == 200:
            with open("summary_audio.wav", "wb") as audio_file:
                audio_file.write(response.content)
            print(f"Audio summary saved as: summary_audio.wav")
            return "summary_audio.wav"
        else:
            print(f"Error generating TTS audio: {response.status_code}")
            return None
    except Exception as e:
        print(f"Error generating TTS audio: {str(e)}")
        return None

# ✅ Full Pipeline Execution
def main():
    video_url = input("Please enter the YouTube video URL: ")  # User input for YouTube URL

    # Step 1: Download Audio from YouTube
    audio_file = download_audio(video_url)
    if audio_file:
        # Step 2: Get Transcription from Hugging Face
        transcript = get_transcription(audio_file)
        if transcript:
            print("Transcript:\n", transcript)

            # Step 3: Summarize the Transcript using Hugging Face Model
            summary = summarize_text(transcript)
            if summary:
                print("\nSummary:\n", summary)

                # Step 4: Generate TTS Audio for the Summary
                tts_audio = generate_tts_audio(summary)
                if tts_audio:
                    print(f"Text-to-Speech audio saved at: {tts_audio}")
                else:
                    print("Failed to generate TTS audio.")
            else:
                print("Failed to summarize transcript.")
        else:
            print("Failed to transcribe the audio.")
    else:
        print("Failed to download audio from the video.")

if __name__ == "__main__":
    main()