I believe Discord also uses Opus codec IIRC. But wonderful to see the codec in a fully open-source project.

Added noise suppression would also be great, especially a FOSS-based one (Discord uses AI-based noise suppression from Krisp [0] -- a Berkeley, CA based for-profit startup).

I'm curious if there are any FOSS community projects in this noise suppression space. Given the rapid rise in WFH culture post covid-19, I can see an explosion in demand for s/w products in this space.

[0] https://krisp.ai/

Except via discord, the quality is throttled and a bunch of random decisions are made about what to do with your audio outside of the codec.

Drives me up the walls when I mention discords awful quality and people just say "I don't know what you're talking about they have the same codec".

It's only a piece of the puzzle. Not to mention the constant robotting because some of their servers are melting or something.

Host your own voice comms :) Even the cheapest 2€ a month box can handle a decent sized party with no issues.

Hosting a Mumble server is the easy part

Convincing everyone else involved to switch over to it however is a near-insurmountable battle.

Many years ago I was playing on a friends Minecraft server, while in a then Skype call with some friends on the server

One user couldn't get Skype to work on their Arch Linux install, and proceeded to spend almost an hour badgering us and whining to switch to Mumble so they could "join the fun"

From our perspective, shifting our entire group conversation over to a brand new voice chat system just for one person seemed like an absolute joke. So they were left wanting

It also doesn't help that Mumble has a very bad UI for the average user.

I love it, because I'm an engineer and know what all those buttons and sliders do. My friends? they didn't like it at all. Thus making the switchover even harder.

This issue has been mentioned in the past already, and I think Mumble devs acknowledged this in the past... having a well polished UI is so important for user adoption! Just throwing all technical choices to the user's face is rarely the right solution.

Does it still badger you about certificate generation on first launch?

I used it with my friend in 2015 as a sort of galactic walkie-talkie on our phones -- he was in another country but we'd just walk around outside and talk to each other. It felt pretty futuristic at the time.

After a few minutes of continuous audio transmission though there would be a significant buildup of latency, so we developed the habit of saying "pineapple" to each other to measure the lag. Best record was 12 seconds.

Yes it does. Or it did circa 2019 when I last logged in for the first time.

Nice to see you had been having fun with it. Skype however had been there since a lot earlier, right? I guess that having to exolicitly "call" your friends removed a bit of that "walkie-talkie" feeling, but still...

What you did was not optimal. You had the opportunity to switch to a better platform and a strong incentive to do so (one member could not participate without it) you should have put the necessary effort to transition. It would have made your community stronger. Generally speaking, the policy of "none left behind" the most powerful bond-building tool available.

This is not the case if setting up and using Mumble would have caused net more friction for the rest of the community. See all the comments about needing to perform constant troubleshooting for mumble users.

You can't force peers to install software and take the time to train themselves on the new software like you can with an employee.

I have long left the online gaming community but this was no issue 10 years ago. We all switched between TeamSpeak 2, 3, ventrillo and mumble seemlessly.

The issue is that Discord's audio processing is focused on mitigating the impact of the worst audio setups, not preserving the quality of the best, and I don't even think that's the wrong focus given their casual target market

I'd believe this if they didn't offer higher quality as a paid feature.

And why is there no passthrough mode setting for voice channels?

Plenty of people out there with decent usb mics and a cheap little dac/amp setup.

A pass-through mode where we disable all audio processing is on the roadmap, for those with really good setups that don't need the post-processing :)

discord has terrible audio and ((non paid) video). I still use teamspeak3 these days with some of my friends being on ts5.

I always get angry when companies use it for commentary on esports because its just that damn awful

Yup. I've done this too. https://youtu.be/eUQKOPyiANc

It's just so bad in discord I don't know what they are doing

Not to mention their internal copy of the audio.

> But wonderful to see the codec in a fully open-source project.

I know that a few open source games use it for in-game chat. Empty Epsilon [1], and my own Space Nerds in Space [2] both use it for this, and there are probably others.

[1] https://daid.github.io/EmptyEpsilon/ [2] https://smcameron.github.io/space-nerds-in-space/

Mumble has RNNoise.

Speaking of Krisp, I’m curious about other people’s experiences using their noise suppression with Discord.

I personally had it enabled (by accident) for over a week, and during that week I noticed people were either ignoring me or not hearing me. I discovered it was the latter when I disabled the noise suppression option. Krisp was suppressing my voice in addition to noise.

Other people I know who’ve tried it have had similar results, so I’m wondering if we’re using it wrong or if Krisp just isn’t worth using at this point.

I've ran into the same problems. My voice gets consistently muted with it on. I've been able to get it to sort of work by manually setting input sensitivity but even then I still get muted at times.

Most calls on the web are using it. It’s the webrtc default

Right. As you would desire and hopefully expect, even if it might be technically possible to squeeze out slightly better compression with some non-free technology, it's very attractive for everybody to agree on the free things instead.

The FSF even offers this as one of the few reasons you wouldn't use Copyleft. It actually would be better for society if everybody copy-pastes the Opus code and refuses to let you see their source code, than for them to use a proprietary codec instead.

I used to work on 3D language learning apps in VR. The voice audio chat was always a problem until I started using Opus. Combine Opus with 3D audio, and it sounds very lifelike and clear, almost as if you are in the same room as the other person!

I'm pretty sure it used to use Speex

According to a 2019 pull request[1]:

> Speex was the first audio codec we used, until it was superseded by CELT in 2009 (Mumble 1.2.0), which in turn was superseded by Opus in 2011 (Mumble 1.2.4).

> We retained support for [Speex] for 10 years and we believe all active servers are currently using Opus.

1: https://github.com/mumble-voip/mumble/pull/3812

Isn't that Ventrilo?

Mumble did use Speex too. I4ve heard the "This is an example of Speex" voice line spinning around my head too many times every time I set up mumble.

Stadia uses Opus as well