SpeechSynthesis API onend callback not working

Question

I'm using the Speech Synthesis API on Google Chrome v34.0.1847.131. The API is implemented in Chrome starting in v33.

The text-to-speech works for the most part, except when assigning a callback to onend. For instance, the following code:

var message = window.SpeechSynthesisUtterance("Hello world!");
message.onend = function(event) {
    console.log('Finished in ' + event.elapsedTime + ' seconds.');
};
window.speechSynthesis.speak(message);

will sometimes call onend and sometimes not call it. The timing appears to be completely off. When it does get called, the printed elapsedTime is always some epoch time like 1399237888.

Answer 1

According to this comment on the bug mentioned in the answer from Kevin Hakanson, it might be a problem with garbage collection. Storing the utterance in a variable before calling speak seems to do the trick:

window.utterances = [];
var utterance = new SpeechSynthesisUtterance( 'hello' );
utterances.push( utterance );
speechSynthesis.speak( utterance );

Answer 2

While this is how I found it to make it work, I am not sure if this is the right behavior....

First don't call the speak function it right away, use callback.

2nd, to get time use timeStamp instead of elapsedTime. You could have just used performance.now() as well.

var btn = document.getElementById('btn');
speechSynthesis.cancel()
var u = new SpeechSynthesisUtterance();
u.text = "This text was changed from the original demo.";

var t;
u.onstart = function (event) {
    t = event.timeStamp;
    console.log(t);
};

u.onend = function (event) {
    t = event.timeStamp - t;
    console.log(event.timeStamp);
    console.log((t / 1000) + " seconds");
};

btn.onclick = function () {speechSynthesis.speak(u);};

Demo: http://jsfiddle.net/QYw6b/2/

you get time passed and both events fired for sure.

Answer 3

You could use the EventListener for start and end like I did for Speakerbot (http://www.speakerbot.de/).

Here my face changes while words are spoken.

newUtt = new SpeechSynthesisUtterance();

newUtt.addEventListener('start', function () {
     console.log('started');
})

newUtt.addEventListener('end', function () {
     console.log('stopped');
})

Answer 4

I found both solutions suggested here not working in an app I just wrote. The only solution I could come up with is a (sort of) busy waiting:

function speak( text, onend ) {
  window.speechSynthesis.cancel();
  var ssu = new SpeechSynthesisUtterance( text );
  window.speechSynthesis.speak( ssu );
  function _wait() {
    if ( ! window.speechSynthesis.speaking ) {
      onend();
      return;
    }
    window.setTimeout( _wait, 200 );
  }
  _wait();
}

you can find a complete example in this codepen

Answer 5

This looks similar to a Chromium bug reported on Jul 12, 2015.

Issue 509488: Web Speech API: 'end' event of SpeechSynthesisUtterance object is not dispatched sometimes

Answer 6

print the utterance before speak seems working... If I remove the console, this issue will happen, don't know why

console.log("utterance", utterThis);
synth.speak(utterThis);

Answer 7

I also found the only way to make this work reliably is to use .cance. I use a 17 second timeout. All of my recordings are under 20 seconds so this works for me.

utterance.onstart = function (event) {
setTimeout(function(){window.speechSynthesis.cancel();},17000);
};

Before I would run into this problem once every 8-10 messages it attempted. Once I added .cancel it seems to always work. I also call set timeout when invoking.

setTimeout(function(){window.speechSynthesis.speak(utterance);},100);

Answer 8

The only thing that worked for me was to avoid voices with localService being true. Those voices never fired onend while the other voices (localService being false) did fire onend.

Answer 9

Also, on Chrome (but not Safari), the callback is never called if you try to speak an empty string.