JavaScript speechSynthesis.speak() without user activation is no longer allowed since M71

This is part of Chrome's new policies regarding making sound from web-pages.
You simply need your user to provide an user-gesture (for which you can find a list here) during the lifetime of the parent document (i.e the event may long be dead, as long as the user ever interacted with the page).

Note that these events can even traverse frames, so for instance, in StackOverflow, the simple fact that you do have to click on the "Run" button will make the inner frame allowed to execute this code:

const ut = new SpeechSynthesisUtterance('No warning should arise');
speechSynthesis.speak(ut);

And in your code, you simply have to provide some kind of an UI that will ensure your users have interacted with the page before you call this method (e.g a button / toggle will do perfectly).