/transcribe

Convert images or audio files to plain text on the command line

Primary LanguagePythonMIT LicenseMIT

Transcribe

  1. verb - To convert a representation of language to another representation.
  2. noun - Command line application to extract plain text from images and audio files, written in Python.

OCR

To extract the text from an image:

$ transcribe ocr /path/to/image.jpg

Speech

To extract the text from an audio file:

$ transcribe speech /path/to/audio.mp3

Installation

Google Cloud setup

Transcribe uses Google's cloud services to perform the text extraction, that means you have to setup a Google cloud project. Below is roughly how I did that...

I went to The console

Selected Project in the top left corner (to the right of Google Cloud Platform) and then create a new project, wait a bit for it to create the project and switch to it (this took like 20 seconds while I was trying to figure out what the heck was going on). Then activate the vision and speech apis (I think storage is automatically activated) and then select Credentials, and create an api key. More auth info here.

Turns out I also needed to create a service json file. To do this I needed to go to the dashboard for the project and then click Create credentials and choose Service account key.

Then add this to your .bash_profile or something similar:

export GOOGLE_APPLICATION_CREDENTIALS="/path/to/keyfile.json"

ffmpeg

If you're on a Mac and you use Homebrew, you can install ffmpeg like:

$ brew install --with-opus ffmpeg

If you're not on a Mac or don't use Homebrew you're on your own.

Transcribe

Install using pip:

$ pip install transcribe

or the latest and greatest:

$ pip install "git+https://github.com/Jaymon/transcribe#egg=transcribe"