Use timeouts: Android Edge speechSynthesis.speak() could get stuck in speaking

[Falcosc]

I don't have a demo, but I can create one based on this revision https://github.com/Falcosc/PSVR2-Twitch/tree/b07c03296f8073309820d2554e1ec71cdf33caa2

It took around 30 minutes of repeating a text every minute. I can't give more detailed information about how to reproduce. But you can download this revision, open index.html and connect to twitch and click start.
I don't know if it is related to fetch requests or if it happens without them too.

What happens?

• speak does not trigger execute SpeechSynthesisUtterance onEnd
• if you execute speak again, it does not start, it goes into pending
• result is SpeechSynthesis state pending=true and speaking=true

Workaround:

• execute speechSynthesis.cancel() and take care about your pending SpeechSynthesisUtterance objects
• use a timeout like setTimout promise in a Promise.race

Maybe it is already handled in EasySpeech, they do an speechSynthesis.pause() and speechSynthesis.resume() from time to time. But maybe it is not, so make sure to guard your speak requests with timeouts if you deal with a known amount of text.

[Falcosc]

The test was done with an empty chat room, so there were no parallel speak requests, no speechSynthesis status checks. Basically best case scenario, no interaction with speechSynthesis at all. Just sending a new text every minute.

Sometimes it took more than 30 min. So difficult to say which edge versions are effected since I don't do such long test sessions in my current project phase.