Reads arXiv papers using Text-to-Speech. Uses the Facebook model facebook/fastspeech2-en-ljspeech.

Install these requirements:

pip install transformers
pip install g2p_en
pip install beautifulsoup4
pip install arxiv-downloader
pip install fairseq

The script also requires pandoc. If you are running on a Mac with Homebrew installed, you can install pandoc with:

brew install pandoc

If you do not run a Mac with Homebrew - which is a reason for regret - follow the installation instructions for your OS.

If you want to convert "Attention is all you need" to speech, you can run the following command:

python convert.py "1706.03762"

Better use the quotes around the arXiv ID.

Feel free to add more features and improve things. Pull requests are more than welcome. And do not hesitate GitHub issues.

Have fun!