Whisper ASR Transcription Project

Overview

This project utilizes the Whisper Automatic Speech Recognition (ASR) model developed by OpenAI to transcribe audio from both a locally hosted video and a YouTube video. The transcription results are saved as text files for further analysis.

Dependencies

• OpenAI Whisper: A powerful ASR model for transcribing speech.
• pytube: A lightweight, dependency-free Python library to download YouTube videos.
• moviepy: A video editing library for Python.
• ffmpeg: A multimedia framework to handle audio and video processing.

Usage

1. Install the required dependencies:

   pip install -U openai-whisper
   pip install ffmpeg
   pip install pytube
   pip install moviepy

2. Load the Whisper ASR model:

   import whisper
   model = whisper.load_model("tiny")

3. Run the provided Python script:

   • Local Video Transcription:

     from moviepy.editor import VideoFileClip
     video = VideoFileClip("video.mp4")
     video.audio.write_audiofile("myaudio.mp3")
     
     result_local = model.transcribe("myaudio.mp3", fp16=False)
     with open("mySound_local.txt", "w") as file:
         file.write(result_local["text"])
     print(result_local["text"])

   • YouTube Video Transcription:

     from pytube import YouTube
     yt = YouTube('https://www.youtube.com/watch?v=Aq92xxwYwSU')
     
     stream = yt.streams.get_highest_resolution()
     video_path = stream.download()
     
     video_youtube = VideoFileClip(video_path)
     video_youtube.audio.write_audiofile("myaudio.mp3")
     
     result_youtube = model.transcribe("myaudio.mp3", fp16=False)
     with open("mySound_youtube.txt", "w") as file:
         file.write(result_youtube["text"])
     print(result_youtube["text"])

4. Analyze the transcriptions saved in the generated text files (mySound_local.txt and mySound_youtube.txt).