Skip to content

Latest commit

 

History

History
24 lines (17 loc) · 1.41 KB

README.md

File metadata and controls

24 lines (17 loc) · 1.41 KB

HearPapers - Listen to Papers

The code convert_to_audiobook.py allows you to pass in the pdf file path and the path to store the audiobook to and it will save a .mp3 file to the path specified.

Running python3 convert_to_audiobook.py -h returns:

usage: convert_to_audiobook.py [-h] --paper PAPER --path PATH

optional arguments:
  -h, --help     show this help message and exit
  --paper PAPER  The path of the pdf paper you want to convert to audiobook
  --path PATH    The save path - to save the audiobook

Note: This is currently tested only for CVPR papers, I am not sure about others since the text extracting follows the CVPR style guide.

Results

For the initial experimentation, I converted the paper Stacked Attention Networks for Image Question Answering from CVPR 2016 to an audiobook. The PDF link of the paper: PDF PAPER. The converted audiobook: LISTEN HERE.

Notes

I used gTTS and therefore the voice is a bit robotic, I plan on using WaveNet in the future and also (smart) inclusion of sentences from the pdf in the audiobook. Currently the only heuristic is if the character length of a sentence is more than 20, we include in the audiobook. The CVPR paper results are good. Equations are a bit tricky, will need to figure that out in the future.