PDF2Audiobook

Check out my blog for step-by-step guide and learning how this tool works.

And see the original articles (Kazunori, Morkowitz) and videos (Kazunori, Markowitz) for reference.

Main process

1. Set ANNOTATION_MODE = True (/functions) for generating annotation data
2. By default, the pdfminer.six is used to extract the content of pdf files, you can set NO_OCR=False to use Google OCR instead.
3. Use annotation tool (/apps-script) to create training data
4. Train a AutoML Tables model
5. Set ANNOTATION_MODE = False for generating mp3 files

Deploy with command line

You can operate entirely in the web editors, but if you prefer registering and deploying with command line, use the following command. You need to create the bucket beforehand as a workspace for PDF2Audiobook.

gcloud functions deploy <FUNCTION_NAME> --runtime python37 --trigger-bucket <BUCKET> --memory=2048MB --timeout=540