ttsreader

ttsreader is a Text to Speech wrapper, reader / player and helpers for the web-speech-api speech synthesis

See published npm package https://www.npmjs.com/package/ttsreader

It is the engine behind the popular ttsreader.com online text & website reader , as well as:

• Audactive - voice enabled learning resources
• Speechnotes' (online dictation notepad) proof read capability
• Speechnlogger's speak out loud

Import:

To use in your own html / js code use one of the following methods:

• <script src="https://unpkg.com/ttsreader/build/bundle.min.js"></script>
• Or - download the following built file /build/bundle.min.js and use directly in your code <script src="bundle.min.js"></script>
• Or - for Node based projects: npm i ttsreader

How to use?

See 'test.html' for a complete example.

The tts engine itself:

let tts = wsGlobals.TtsEngine; OR import {TtsEngine} from "tts-engine"

See test.html for real example on use. test.html

Tests Documentation

Since there are so many different platfrom-browser combos, we have to carefully test on each. Documentation bout it is here: TESTS.md

Showcases

This is used on the following sites:

• Speechnotes - Lightweight Speech Recognizing Notepad
• TTSReader - Online Text To Speech Reader

TODOs:

• Utils to 'cut' large texts into smaller sections.
• Generic Player with UI, with options to play
  • text
  • element and its contents

How to contribute & publish updates to this package?

• This package is published both on GitHub and on npmjs:
  • https://www.npmjs.com/package/ttsreader
  • https://github.com/RonenR/ttsreader
• To publish an updated version - simply run npm publish, it will commit & push updates both to github and npm.

Important practical non-obvious lessons we learned about tts:

• You cannot send too many chars to REMOTE tts voice. It jams. Probably - the whole text is processed at once, and some mp3 is generated on the server?
  • 38,000 is too much. It basically never starts.
  • Local voices (at least on mac) worked well.
• With Google's voices - they will use the correct voice only sometimes. It's not consistent at all! (On Mac.) // TODO: Test on Windows!
  • Solution: If you send a single utterance - it reads with a consistent voice.
  • Problem: if they get the wrong voice - it will mess up reading for a large portion of the text...
  • Sent google a bug report: https://bugs.chromium.org/p/chromium/issues/detail?id=1344469
• onboundary does not work with Google voices. So - we have to send small chunks anyhow...
• Of course - the famous google voice terminating itself after some circa 15 secs...