I know someone missing a lot of the high end, and he has really enjoyed the pandemic because all of his social interactions are online, where everyone is wearing headphones, and everyone is suffering from the same handicap:

No conferencing software that I'm aware of has solved the cocktail party effect. Since nobody can triangulate sounds properly, nobody can filter out side conversations or external noises. Everyone has to take turns to be understood, and if they don't, the guy with hearing aids doesn't get singled out by having to bring it up.

For a hackathon last year, I patched Jitsi to play each person's audio from a different angle using the web audio spatialization APIs[1], and it really does help the cocktail party effect tremendously- people laughing or talking over each other (extremely common in physical conversations) don't make it impossible to understand the other person like when everything is mixed to a mono channel as it is with most video conferencing software.

1. https://developer.mozilla.org/en-US/docs/Web/API/Web_Audio_A...

That's an awesome idea & project, but mightn't it feel jarringly like being surrounded by people talking at you?

I think I'd at least want them all in the, say, central 160deg in front of me. At which point I wonder how good the 'spatialisation' is, how many people could be meaningfully supported.

Because you can't just have alternate speakers spaced out for maximal distinction - the same speaker suddenly coming from somewhere else would be even more jarring.

Really cool though.

Apple just announced this as a feature for FaceTime at WWDC (though not out until the fall, which seems poorly timed). I’ve wondered for the past year why no major conferencing software has implemented that.

I heard about VocalZoom at the Embedded Online Conference, who are trying a hardware approach to solving the cocktail party effect.

Basically they use lasers and directional microphones. Seems like it could work for a conference call in a meeting room, but not sure how helpful it would be for a party.

https://patents.google.com/patent/US8286493

I'm more interested in the SRAY: a directional loudspeaker, which could avoid the need for so much ambient background noise. Also useful for alarm clocks, I suppose.

https://www.indiegogo.com/projects/sray-world-s-smallest-dir...

Mumble does have positional audio but it has to have a game/input source that supports reporting the position of the player - always wanted to test it one day.

I'm deaf on one side, so issues like this tend to be topics that I've already given a lot of thought to.

A counter I employ to the cocktail-party effect is to pitch my voice very low. Sometimes the person I'm talking to picks up on this (consciously or not!) and imitates my low pitch. Despite others talking over us, it's easier to focus in on a lower pitch and let the brain filter out the noise.

It can be done with a normal PC but I loved meeting people in a VR environment where you could walk around and form groups and you only heard audio from nearby people. They also had a movie theater with a movie everyone could watch at the same time.

But any interface where you can move a character in a world would be enough to make this work. It’s lovely because you can drop in and listen to a convo and walk away to another if you’d like.

You might enjoy https://gather.town :D

Ooh looks great!

Pretty much all the VR ones do this, as you have a spacial representation to work from, from what I've seen. And there have been some recent-ish things like https://gather.town/ that should be capable as well, though I'm not sure if they do. But yeah, Zoom doesn't.

Cool thing about VR is when you move your head the position of the speaker stays exactly where it is in relative space thanks to motion tracking which lets you essentially turn towards people you wanna listen/talk to. At some point we need some sort of positional awareness in our headphones to add some resemblance of that.

You could use camera head tracking for that.

I wonder if this would apply to mixing in a frequency that would result in an audible heterodyne mix (think like active noice cancelling but using a waveform that is designed to downconvert vs cancel).

I do believe there is a method to use this to give blind folks an ambient doppler tone from objects in their vicininity. I just don't have the chops or time to try it out.

I wonder if you could build some sort of deliberate reflector into a microphone housing to generate comb filtering that an analytic technique (or ML) could derive spatial data from. I know in the limit this is possible, since there's dummies used for binaural recording or testing earbuds, but could an intentionally designed mechanism accomplish the same thing in a smaller package?

It'd be neat if you could then encode the spatial information and use an HRTF on the destination receiver to appropriately place the sound in the listener's soundstage.

I’d love to hear an audio example that demonstrates this effect.