1606759 - Google Meet incoming audio garbled

Status: RESOLVED FIXED

(Core :: WebRTC: Audio/Video, defect, P2)

Description

[Kelsey Gilbert [:jgilbert]]

Possibly related to bug 1518638.

I can reliably repro this on at least one Windows machine.

This is a regression from around 2019Q4. Maybe I can find a regression window.

Is there anything (about:support?) I can offer to sanity check things?

Comment 1

[Kelsey Gilbert [:jgilbert]]

This makes me need to use Chrome.

Comment 2

[Dan Minor [:dminor]]

Based on the timing, I wonder if Bug 1586370 is the culprit. A regression window would be very helpful.
Paul, could you please have an initial look at this? I can follow up if it turns out to be a webrtc.org problem.

Comment 3

[Paul Adenot (:padenot)]

(In reply to Jeff Gilbert [:jgilbert] from comment #0)

Is there anything (about:support?) I can offer to sanity check things?

Please copy here the content of the Media section in about:support, thanks!

It would be helpful to have an audio recording of this, if possible.

Comment 4

[Kelsey Gilbert [:jgilbert]]

Attachment: about:support Media

Comment 5

[Kelsey Gilbert [:jgilbert]]

Attachment: never-gonna-give-you-up-screencap.mp4

Comment 6

[Paul Adenot (:padenot)]

You have lots of devices. It looks like your output is an optical output (SPDIF), is that right?

Would you mind doing what you do in the video, with a bit more logging (this is going to output lots of data hence the MOZ_LOG_FILE):

MOZ_LOG=cubeb:5 MOZ_LOG_FILE=something.log

I'd like to understand how your audio output device is configured. Here, analyzing the waveform of the recorded audio, it has a valid signal interleaved with noise that is almost silence, at roughly equal duration.

Was the audio playing through what the name of the file suggests? I think I can reconstruct it somewhat in audacity by trimming the silences, and applying a resampling pass, but I'm not sure. This hints at what the possible bug in Gecko is.

A regression window will surely help, as I have at least a couple different candidates for the cause of this, that would fit the approximate range your mentioned in the description.

Comment 7

[Kelsey Gilbert [:jgilbert]]

That is indeed a clip from Never Gonna Give You Up. I hear it as somehow both sslloowweedd ddoowwnn audio, but somehow also playing back on time, 1:1. All other browser audio (and videos) work fine.

My main output is SPDIF optical, yes.

I'll try for a log.

Comment 8

[Kelsey Gilbert [:jgilbert]]

Attachment: cubeb.log

Comment 9

[Kelsey Gilbert [:jgilbert]]

New data: Disabling microphone access/permission on the receiving machine causes its audio output to revert to normal!

Comment 10

[Paul Adenot (:padenot)]

Jeff, would you mind trying again? We've changed a few things recently that might have fixed this, specifically on Windows.

In particular, I see in your log that your computer is able to use IAudioClient3, that we had enabled by mistake: this is a path that should have not been reached.

Comment 11

[Kelsey Gilbert [:jgilbert]]

Seems fixed in my STR! Thanks!