Search code examples
tensorflowspeech-recognitionspeech-to-textgoogle-speech-api

Offline Speech Recognition in browser


I am working on a product that need to take inputs from user and do certain actions based on it. We have implemented it with a chat box via typing and it is serving our purpose. For the future releases we want to add voice recognition to the chat window. We thought of using

window.speechRecognition() || window.webkitSpeechRecognition()

but we came to know that the functionally available in browsers use Google's Cloud Speech API. As we deal with very sensitive information of users this will be security issue. Is there any other alternatives for implementing the speech recognition that works in any browsers.


Solution

  • Apparently PocketSphinx.js is the only available way to go as of now. It's an open-source speech-to-text engine that supports English but not many languages beyond that.

    Github:

    However, if you want to run your code on a single instance of an Android device (e.g. a device displayed somewhere in a public area), you can use "Download offline voice recognition language" in mobile Chrome's settings. There is no such option for the desktop browser.