mdn
api.SpeechRecognition.continuous - false unsupported in Safari
What type of issue is this?
Browser bug (a bug with a feature that may impact site compatibility)
What information was incorrect, unhelpful, or incomplete?
Safari does not honor SpeechRecognition.continuous being set to false. It does not stop listening after it stops. See my StackOverflow post: https://stackoverflow.com/questions/75498609/safari-webkitspeechrecognition-continuous-bug
What browsers does this problem apply to, if applicable?
No response
What did you expect to see?
Detailing this bug in Safari
Did you test this? If so, how?
Read: https://stackoverflow.com/questions/75498609/safari-webkitspeechrecognition-continuous-bug
Can you link to any release notes, bugs, pull requests, or MDN pages related to this?
https://stackoverflow.com/questions/75498609/safari-webkitspeechrecognition-continuous-bug
Do you have anything more you want to share?
No response
MDN URL
https://developer.mozilla.org/en-US/docs/Web/API/SpeechRecognition/continuous
MDN metadata
MDN page report details
- Query:
api.SpeechRecognition.continuous - Report started: 2023-03-09T13:50:26.194Z
Oof, I ran into this as well. Thank you for your StackOverflow post.
Safari does seem to respond to the continuous flag; it only emits an end event when continuous is set to false. I tested on Safari 16.5.2 on MacOS 11.7.10.
Here's a standalone HTML file that reproduces the behavior that results are still emitted even though continuous has not been set (and so defaults to false).
<!DOCTYPE html>
<html>
<head>
<title>Safari Speech Recognition Is Always Continuous</title>
</head>
<body>
<p>Click anywhere to start speech recognition. The expected behavior is that the browser stops listening after the `end` event fires, but it seems that Safari continues to listen and return results.</p>
</body>
<script>
document.onclick = async () => {
const SpeechRecognition = window.SpeechRecognition || window.webkitSpeechRecognition;
const recognition = new SpeechRecognition();
recognition.onend = (event) => {
document.body.innerHTML += `<p>End</p>`;
};
recognition.onresult = (event) => {
document.body.innerHTML += `<p>Result: ${event.results[event.results.length - 1][0].transcript}</p>`;
};
document.body.innerHTML = '<p>Listening...</p>';
recognition.start();
};
</script>
</html>
@tsegreto @paulkernfeld Can you confirm whether you still experience the issue in the latest Safari version? (I cannot reproduce it locally with Safari 18.2.)
It was probably fixed with https://github.com/WebKit/WebKit/commit/68c30e63bd2664d632c899686f369678e11685a7 (https://bugs.webkit.org/show_bug.cgi?id=252936) in WebKit 616.1.4 between Safari 16.6/615.3.12 and Safari 17/616.1.27.
Sorry, I'm on Intel so I think I probably can't upgrade to that version of Safari (I'm on 16.6.1 now).
Based on current evidence in this issue, we should update the Safari statement for continuous, splitting it up into [{ version_added: "14.1", version_removed: "17", partial_implementation: true, notes: "Returns multiple results when set to `false`." }, { version_added: "17" }]:
https://github.com/mdn/browser-compat-data/blob/7d5252f3ad23908e699751e36163825fd9125e61/api/SpeechRecognition.json#L222-L224