Hearing own voice - feedback filtered in Talk/OS/Hardware/...?


Some people don’t use headphones but speakers & microphone (usb or 3.5mm plug).

Sometimes that means other people hear themselves, sometimes that does not happen. Looks like somewhere it is recognized that a sound pattern that was just played via speakers should not be recorded/forwarded.

Where does the filtering/suppression happen?

  1. hardware level/driver of the microphone?
  2. operating system level? (is it different for Win/Mac/Linux?)
  3. in Talk (in local browser java script? in High Performance Backend?)

This is not about a specific setup. I want to understand the general situation. If it depends on certain circumstances, please explain the different possibilities.

Thank you =)

So I’m not sure, might be that 1 + 2 also do it depending on the amount of money spent.
3 is not it, but the browser does some echo-cancelation itself.

Thank you. I am looking for someone who really knows. Hopefully someone will reply =)

Joas Schilling via Nextcloud community <noreply@nextcloud.com> hat am May 12, 2021 um 9:18 AM geschrieben:

@nickvergessen I just saw that you are Talk lead. Leads don’t need to know everything of course, but can you point me to a person/resource that will have the answer?

I’m not only the lead but also the backend dev. It’s about the best answer you will get.
As per Option to disable auto gain control / audio processing for practicing music · Issue #3252 · nextcloud/spreed · GitHub the browsers do cancel and some people are annoyed by it. Some professional headset/audio devices will cancel additionally on their own.
The otherthing is auto gaining of the voice level in Chrome which could be part of the problem.

Thank you for your expertise Joas!

Then I assume Safari does echo cancelling, too.

And what about operating system level? (is it different for Win/Mac/Linux?) They hopefully don’t do echo cancelling if the browsers already do it…

Is Chromium better than Firefox at cancelling? (on Linux)