doc2audiobook.py

Extract text from a document (textract) and convert it into a natural sounding synthesised speech (Cloud Text-To-Speech), which is able to leverage Deepminds Wavenet models.

Example

Input Output

Available source formats (from textract)

• .csv
• .doc
• .docx
• .eml
• .epub
• .gif
• .jpg and .jpeg
• .json
• .html and .htm
• .mp3
• .msg
• .odt
• .ogg
• .pdf
• .png
• .pptx
• .ps
• .rtf
• .tiff
• .txt
• .wav
• .xlsx
• .xls

Prerequisites

GCP

1. Select or create a Google Cloud Platform project.
2. Enable billing for your project.
3. Enable the Cloud Text-to-Speech API.
4. Setup Authentication using a Service Account.

Host Machine

1. Docker
2. /doc2audiobook/data/input: directory to hold all input files.
3. /doc2audiobook/data/output: directory to store all output files.
4. /doc2audiobook/.secrets/client_secret.json: GCP authentication token.

Build

git clone git@github.com:danthelion/doc2audiobook.git
cd doc2audiobook
docker build -t doc2audiobook .

Run

Make sure to put your documents in the folder that is mapped to /data before running!

List available voices

docker run \
-v /doc2audiobook/data:/data:rw \
-v /doc2audiobook/.secrets/client_secret.json:/.secrets/client_secret.json:ro \
doc2audiobook -list-voices

Convert all documents in the mapped input folder to audiobooks using the en-GB-Standard-C voice.

docker run \
-v /doc2audiobook/data:/data:rw \
-v /doc2audiobook/.secrets/client_secret.json:/.secrets/client_secret.json:ro \
doc2audiobook --voice en-GB-Standard-C

Convert a single document in the mapped input folder to an audiobook using the en-GB-Standard-C voice.

docker run \
-v /doc2audiobook/data:/data:rw \
-v /doc2audiobook/.secrets/client_secret.json:/.secrets/client_secret.json:ro \
doc2audiobook --voice en-GB-Standard-C --input test_input.txt