html5cat / voice-google-speechtotext

Demo application of Google Cloud Speech recognition

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Google Cloud Speech Transcription Demo

This is a demo applicaiton of using Google Cloud Speech API with Nexmo Websockets to perform realtime transcription in multiple languages.

Google Cloud Speech API

You'll need to signup for the Speech API as part of Google Cloud, start here https://console.cloud.google.com/launcher/details/google/speech.googleapis.com

Prerequisites

The application is built in node you will need the following packages:

  • dotenv
  • websocket
  • httpdispatcher
  • url
  • @google-cloud/speech
  • nexmo

These are all specified in the package.json

Once you have enabled the API you'll need to create a project and get a set of credentials, these will be downloaded as a JSON file, save that file as google_creds.json in this project.

Nexmo

You will need to create a Nexmo Voice Applicaiton with the answer url set to [YOURHOSTNAME]/ncco and optionally the event URL set to [YOURHOSTNAME]/event You then need to purchase a number and link that to the applicaiton, you can edit the number that is displayed on the webpage in index.html to match your number.

Configuration

Apart from the Google credentials the only other parameter to configure is the hostname(s) The server needs to listen on 2 separate TCP ports to expose a websocket server for the browser and for the nexmo voice API to connect to, the defaults for this are port 8000 and port 8001. Almost everything is on port 8000 apart from the nexmo websocket connection which uses 8001, you can either run this server using ngrok and your local machine then expose each of the 2 ports under different hostnames or use the same hostname for the machine and just put :8000 & :8001 on the end of the hostnames. These are configured in the .env file.

Running

To start the server execute node ./server.js You can then browse to the root of your hostname and the page should be served, call the number linked to your applicaiton, enter the language code you want to test (e.g. 16 for UK English) and once the language has been confirmed to you start talking. You will see transcription on the right hand side of the page.

About

Demo application of Google Cloud Speech recognition

License:MIT License


Languages

Language:JavaScript 64.4%Language:HTML 35.6%