Skip to content

Latest commit

 

History

History
118 lines (83 loc) · 4.19 KB

README.md

File metadata and controls

118 lines (83 loc) · 4.19 KB
Raw

Audio and Video Transcription Tool

This Python script allows you to transcribe audio and video files using the Wit.ai API. It supports transcription of YouTube videos, local audio files (WAV and MP3), and local video files (MP4, MKV, AVI).

Features

  • Transcribe YouTube videos by providing the video link
  • Transcribe local audio files in WAV or MP3 format
  • Transcribe local video files in MP4, MKV, or AVI format
  • Automatically convert MP3 files to WAV format for transcription
  • Automatically extract audio from video files for transcription
  • Support for multiple languages (requires Wit.ai API key for each language)
  • Generate transcriptions in TXT and SRT formats

Requirements

  • Wit.ai API key for each language you want to transcribe
  • Docker (if using the Docker setup). if not using Docker:
  • Python 3.x
  • ffmpeg
  • yt-dlp
  • tafrigh library (pip install "tafrigh[wit]")
  • ffmpeg-python library
  • pydub (install this if you get an error)

Installation

  1. Clone the repository:
git clone https://github.com/marouane53/transcribe.git

Running with Docker (Recommended)

  1. Build the Docker image:
cd transcribe
docker build -t transcribe-app .
  1. Run the Docker container, mounting the local downloads directory:
docker run -it --rm -v $(pwd)/downloads/:/usr/src/app/downloads transcribe-app

Replace $(pwd)/downloads/ with the path to your local downloads directory you wish to use for saving the transcription results. This ensures that any files downloaded or created by the script will be accessible on your host machine.

Running Locally

  1. Install the required Python packages: 3. Install ffmpeg:

  • For Windows:

    • Download the ffmpeg build from the official website: ffmpeg.org
    • Extract the downloaded archive and add the bin directory to your system's PATH environment variable.

  • For macOS (using Homebrew):

    brew install ffmpeg

  • For Linux (using apt):

    sudo apt update
sudo apt install ffmpeg
  1. Install yt-dlp:

  • For Windows:

    pip install yt-dlp

  • For macOS and Linux:

    sudo pip install yt-dlp
  1. Set up the environment variables:
  • Copy the .env.example file to a new file named .env in the root directory of your project.
  • Open the .env file and replace the placeholder values with your actual Wit.ai API keys and any other necessary configuration values.

Note: The .env file is where your sensitive information such as API keys will be stored. It is crucial to ensure this file is never committed to your version control system. The .gitignore file in this project is already configured to exclude .env.

  1. Set up Wit.ai API keys:
  • Sign up for a Wit.ai account at wit.ai if you don't have one.
  • Create a new app for each language you want to transcribe.
  • Obtain the API key for each language.
  • Open your .env file and fill in your Wit.ai API keys next to the corresponding language variables (e.g., WIT_API_KEY_ENGLISH, WIT_API_KEY_ARABIC, etc.).

Windows Release

For Windows users, a pre-packaged release is available that includes the necessary executable files (ffmpeg and yt-dlp) along with the Python script. You can download the release from here.

The release version eliminates the need for installing ffmpeg and yt-dlp separately, as the executable files are included in the package. Simply extract the ZIP file and run the script directly.

Usage

  1. Run the script: python transcribe.py

  2. Choose whether you want to transcribe a YouTube video or a local file.

  3. Follow the prompts to provide the necessary information (YouTube link or file path, language sign).

  4. The script will download the audio (for YouTube videos), convert the file to WAV format (if necessary), and start the transcription process.

  5. Once the transcription is completed, the generated files (TXT and SRT) will be saved in the same directory as the input file.

Acknowledgments

This project was inspired by and adapted from the Tafrigh project.

License

This project is licensed under the MIT License.