Youtube Diarization using pyannote-audio

This tools download audios from youtube and execute diarization by speaker using pyannote-audio.

Installation

This project has been tested on Ubuntu 20.04.4 LTS. First, you need to install the FFmpeg program using the following command:

$ sudo apt update; apt install ffmpeg

To create a conda environment from scratch, run the following commands

$ conda create -n youtube_diarization python=3.9 pip

Activate the environment:

$ conda activate youtube_diarization

Use the package manager pip to install the requirements.

First, install speechbrain (torch is a requirement):

$ conda install pytorch==1.11.0 torchvision==0.12.0 torchaudio==0.11.0 -c pytorch
$ pip install speechbrain

After, install pyannote-audio:

$ pip install https://github.com/pyannote/pyannote-audio/archive/develop.zip

Finally, install other requirements:

$ pip install -r requirements.txt

Or use the yml file:

$ conda env create -f environment.yml

Configuration

Create a youtube links file, one link for line:

$ cat input/links.txt

Your file should look something like this:

# Yoshua Bengio interview
https://www.youtube.com/watch?v=Tzuhnuo8KQg
# Ian Goodfellow interview
https://www.youtube.com/watch?v=Z6rxFNMGdn0
# Andrew Ng interview
https://www.youtube.com/watch?v=0jspaMLxBig
# Geoffrey Hinton interview
https://www.youtube.com/watch?v=-eyhCTvrEtE

If the line starts with '#' it will be ignored.

Execution

For execution it is necessary to use a configuration file, such as the config.json file:

{
  "youtube_list": "input/links.txt", 
  "videos_folder": "output/"
}

• youtube_list: input filepath containing the youtube links.
• videos_folder: output folder in which downloaded audios/videos will be segmented.

To execution, run the command:

$ python main.py -c config.json

License

Apache 2.0