Spaces:
Sleeping
Sleeping
| 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() | |