kidanger/latex-to-video

Convert a beamer presentation into a video using text-to-speech.

latex-to-video

Installation

System requirements:

• sox
• ffmpeg
• 'convert' from imagemagick
• conda, not mandatory

Download the weights:

$ wget https://github.com/kidanger/latex-to-video/releases/download/v1/data.tar.gz
$ tar xvf data.tar.gz

Tested from a fresh conda with python 3.7:

$ conda create -n tts2 python=3.7
$ conda activate tts2
$ git clone https://github.com/mozilla/TTS
$ cd TTS
$ git checkout b1935c97
$ pip install -r requirements.txt
$ python setup.py install
$ cd ..
$ pip install inflect fire mutagen iio ffmpeg-python

Try the TTS:

$ python tts.py 'bla bla bla' out.wav
$ play out.wav

Try the demo:

$ cd demo
$ latexmk -pdf main.tex
$ python ../process.py main.tex main.pdf main.mp4
$ vlc main.mp4

Installation (docker)

Build the docker image locally:

$ docker build . --tag latex-to-video

Try the demo: (recommended)

$ cd demo
$ latexmk -pdf main.tex
$ docker run --rm -u (id -u) -v (pwd):/data latex-to-video main.tex main.pdf main.mp4
$ vlc main.mp4

Usage

Try on your slides:

$ cd somewhere
$ latexmk main.tex -pdf
$ conda activate tts2
# remove --fast for the final build
$ python /path/to/latex-to-video/process.py main.tex main.pdf main.mp4 --fast
$ vlc main.mp4

syntax:
    %>next        -> go to next slide
    %>pause       -> 500ms pause
    %>shortpause  -> 100ms pause
    %> <text to be speeched>

Warning: do not use sentence that would be too long or the TTS might fail. It works on a line per line basis, so put each sentence in its own line.

Credits

TTS engine by Mozilla

Code and weights of the tts taken from colab

PDF to mp4 conversion by Charles Hessel