Open Speech Recording is a small web application to collect short snippets of speech, and upload them to cloud storage. It's designed to help gather open speech data sets to train machine learning systems.

It's based around a small Flask app that will run on Google App Engine. This serves up a client-side Javascript app that prompts for a series of words, records the audio, and then POSTs the results back to the server.

1. Install pip dependencies with pip install -t lib -r requirements.txt.
2. Edit app.yaml with updated environment variables.

If you have the Google Cloud SDK set up, you should be able to run a local copy with this command:

dev_appserver.py app.yaml

I've often had trouble getting local copies of the app to work with cloud storage, so you may see errors on the final upload stage with this setup. To deploy it to an appspot instance, run this:

gcloud app deploy

Phrases that we want to collect are stored in static/phrases.json. Recordings will be stored in the bucket specified in the app.yaml environment variable, in the path: /$phraseKey_$sessionId_$UUID.ogg.

Thanks to the Mozilla team for the Web Dictaphone sample application that I used as a starting point, Sole for the oscilloscope, and the Flask team for a lovely Python microframework!

Written by Pete Warden, pete@petewarden.com.