Hacker News Reader

> I invite you to test out your own "spatial audio" abilities in your environment, including: * locating sound sources within a few degrees and pointing at them

I haven’t tried the other two, but with a pair of AirPods Pro and a video source that support spatial audio, this is ridiculously easy to do, and 100% accurate. I don’t know what witchcraft they’ve employed, but I can easily pick up my iPhone with my eyes closed based on where the audio is ‘coming from’.

Also, with SA turned on, because it sounds like the audio is coming from outside your head instead of from your headphones, it’s much more enjoyable to listen to even without a bunch of surround sound stuff. Things like dialog sound much more ‘normal’.

You don't have to believe it, it's been working since last October.

I understand your skepticism, but their implementation works and it's amazing.

Is it exactly like real life? Of course not. To address your points:

  * Yes, you can absolutely locate sound sources within a few degrees, it's astonishingly accurate
  * No, you can't really estimate sound distance. It's absolutely coming from outside your head, but all sounds like it's kind of just a generic 5 ft away or something
  * Of course you can't walk around your environment blind with them, it's not trying to

You say the spatial perception isn't real, but it absolutely is. And I have no complaints about distortion -- yes obviously the waveform is different, but I perceive zero perceptual degradation in quality. It doesn't introduce any unwanted artifacts that I can notice, at least.

I highly suggest you try it with AirPods Pro and existing Apple TV+ content.

And remember, it's trying to simulate a real-life surround sound speaker setup. Not simulate real life itself. Movies aren't mastered to sound like real life, they're mastered to give a spatial audio component to the movie.

There's a questionable assumption in the HRTF article, namely that the transfer function is the same (or similar) for different people in different environments. I don't think the same HRTF would sound convincing indoors and outdoors, even. Let alone differences in ear shapes and sizes.

That assumption is baked into almost all the implementations, except for some which take measurements of your ears.

To be fair I haven't tried one of those systems, but few people ever have and few people ever will because:

1. It's a hassle to measure your ears

2. All the problems with distortion wrecking timbre will still be there.

Sony is working on a version that uses an image of your ear to find a better match. They talked about it in a few PS5 talks.

They said they might do that in the future, but who knows. PS5 ships with 5 profiles you can test and see which feels best.

Its out. WMX-100M4s and a tidal/ deezer/ nugs subscription are all you need.

https://electronics.sony.com/360-reality-audio

I bet that personalized HRTF is locally encrypted and can't be reused outside these services. is anyone know for sure?

That's cool. Is it out for PS5 yet?

Sure, but commercial spatial audio is in its infancy. We’re maybe in the 1920’s if you compare it to video. So as the techniques expands and people adjust and both sides gain experience we’ll be able to compensate for the perceived shortcomings by exaggerating effectively to make it seem better than our natural spacial hearing.

Fair warning, any attempts in this area will have to take your individual ear shape into account to be accurate. Our brains have learned how sounds change based on the frequency changes due to our individual ear shapes, and any software that wants to emulate true spatial sound will have to take this into account.

This sounds like a very interesting project, especially if it could be done just by taking pictures of both your ears (I'm not sure if that's sufficient to capture all of the topology.) Generating a sufficiently large dataset to learn how that topology relates to sound perception sounds incredibly expensive though.

Sure, but I'd bet that accounting for a handful of shapes (maybe less than 10) could cover 90% of the differences. We've applied that principle to everything else in the body pretty well so far. It will take a good bit of time to figure out what those shapes might be though.

To get there people have to have a clear perception of what "the gap" is between what is and what's possible -- like the error signal in a negative feedback loop.

In reality people will think a service sounds better if Dr. Dre, Jay-Z or Neil Young is behind it.

If people watch a bad orchestra playing in good clothes or they watch a good orchestra play in bad clothes they'll likely think the well dressed orchestra sounds better.

A world like that just doesn't have the capacity to improve.

Spot on about the orchestra. A chap used to know used to work for Midas and said that visuals affect perception of quality significantly. They had a problem with a digital desk where the UI was running at very slow framerates and the customer complained that it sounded terrible. If they turned around and listened without looking at the screen they thought it sounded absolutely fine - merely seeing the flickery slow framerates made them believe it SOUNDED terrible.

And you're right about the inability to improve. Try reading up above where people are arguing that they can't hear the difference between a WAV and a 128 kbps MP3....!! Either their hearing is absolutely shot or they are listening to silence.

I think you're being overly negative. Is it as good as real life? Of course not, but it's not terrible as you're making out. Try Google Resonance: https://cdn.rawgit.com/resonance-audio/resonance-audio-web-s...

Personally I find the horizontal angle localisation excellent - vertical is slightly hit and miss.

But it probably depends on the person, too - HRTF models assume an "average" head

I have made some extremely realistic sound recording by using binaural microphones that are positioned inside my ear canals. Using my own head retains a lot of that sensory information.

loosely related...

  * your ability to sense walls in your environment from the sound of your footsteps reflecting off them and not walk into them

I think you can actually navigate in the dark by clicking your tongue or your fingers and detect obstacles. I know some blind people actually use this echolocation to help them get around. Might be fun to try.