ytws 0.3.0

YouTube Whisper - A YouTube Downloader with Transcription

🎥 YTWS: YouTube Download and Faster-Whisper Subtitle Generation Tool 🔥

YTWS is a tool provides a seamless way to download videos from YouTube and transcribe their audio content. It leverages the capabilities of yt-dlp for efficient video downloading and the faster-whisper for accurate transcription.

⭐ Features

• Easy to Use: Start without any hassle.
• Fast Subtitle Generation: Harness the speed of faster-whisper.
• GPU Acceleration: A simple guide for harnessing GPU power.

⚙ Requirements

ffmpeg

For Python virtual environment (recommended):

conda install ffmpeg

For Windows scoop users:

scoop install ffmpeg

💽 Installation

Install YTWS easily with these commands:

pip install ytws

or

pip install git+https://github.com/faker2048/youtube-whisper.git

🚀 Usage

Downloading and Transcribing YouTube Videos/Audios

To download and transcribe a video, simply run:

❯ ytws dt --help
Usage: ytws dt [OPTIONS]

  Download video/audio from Youtube and generate subtitles.

Options:
  -u, --url TEXT         Youtube URL.  [required]
  -n, --threads INTEGER  Number of threads to use.
  -f, --format TEXT      Download format. See: https://github.com/yt-dlp/yt-
                         dlp#format-selection-examples
  -t, --translate        Translate the subtitles to English.
  -m, --model_name TEXT  Name of the model to use. e.g.  (tiny, tiny.en, base,
                         base.en, small, small.en,          medium, medium.en,
                         large-v1, or large-v2), recommended: (large-v2)
  -r, --model_root TEXT  Root directory for the models.
  --cpu                  Use CPU instead of GPU. This is useful if you do not
                         have a GPU.
  -s, --srt_only         Only generate subtitles. Do not download video.
  -v, --video_only       Only download video. Do not generate subtitles.
  --help                 Show this message and exit.

Example:

ytws dt -u "https://www.youtube.com/watch?v=XXXXXXXX" -m large-v2 # Download and transcribe the video
ytws dt -u "https://www.youtube.com/watch?v=XXXXXXXX" -f "bestvideo+bestaudio/best" 

After running the command, the media file and the generated subtitles will be saved in the same directory as the video file.

Only Downloading YouTube Videos/Audios

To download a video, simply run:

❯ ytws d --help
Usage: ytws d [OPTIONS]

  Download a video/audio from YouTube.

Options:
  -u, --url TEXT         The URL of the YouTube video or the filename to be
                         downloaded.  [required]
  -n, --threads INTEGER  The number of threads to use for downloading. Default
                         is 16.
  -f, --format TEXT      The download format. Options include:
                         'worstaudio[tbr>100]/bestaudio/best' for audio only,
                         'bestvideo[height>=1080]/bestvideo' for video only,
                         'bestvideo+bestaudio/best' for both video and audio
                         (default). Use 'best' for the best quality or specify
                         particular formats.
  --help                 Show this message and exit.

Only Transcribing Videos/Audios

To transcribe a video, simply run:

❯ ytws t --help
Usage: ytws t [OPTIONS]

  Transcribe a video/audio file.

Options:
  -f, --file TEXT        The media file to be transcribed.  [required]
  -t, --translate        Translate the subtitles to English.
  -m, --model_name TEXT  Name of the model to use. e.g.  (tiny, tiny.en, base,
                         base.en, small, small.en,          medium, medium.en,
                         large-v1, or large-v2), recommended: (large-v2)
  -r, --model_root TEXT  Root directory for the models.
  --cpu                  Use CPU instead of GPU. This is useful if you do not
                         have a GPU.
  --help                 Show this message and exit.

Example:

ytws t -f "video.mp4" -m large-v2

After running the command, the generated subtitles will be saved in the same directory as the video file.

🛫 GPU Acceleration

For easy CUDA and cuDNN installation:

• Linux:

conda install cudnn

• Windows:

Although using additional torch libraries may not be ideal, it's a straightforward approach for Windows (compared to Nvidia's official CUDA installation). Any alternative methods or suggestions are welcome!

pip3 install torch --index-url https://download.pytorch.org/whl/cu118

After these settings, YTWS should work with a GPU.

🪄 Available Models

Visit here for more details.

🌟 Contributing to YTWS

We welcome contributions to YTWS! Your support makes a difference.

How to Contribute

• Fork, Clone, Branch: Start by forking and cloning the repository, then work in separate branches.
• Commit: Follow the project standards and commit your changes.
• Pull Request: Submit a concise pull request (PR) to the main branch.

Using Pre-Commit

To maintain code format:

pip install pre-commit
pre-commit install

Pre-commit will guide you through necessary checks and corrections. Feel free to reach out if you have any questions or suggestions!