Hacker News Reader

Comment by aeturnum | Hacker News Reader

It's great to see that graphics continue to improve, but I think the pipelining of reproducing physical space characteristics is most interesting to me. One of the things they mention in the video is using tools to measure how real spaces echo and then reproducing those echo characteristics in virtual spaces. Techniques like that have enormous potential for allowing us to use virtual space and actual space collaboratively. We'll always have games, but I'm excited to see how, in the next twenty or thirty years, technology that started in games starts to allow us to interact and socialize in new ways.

View on HN

Ah, convolution reverb. We've had it for a long time in DSP. Basically what you do is pop a balloon (create an "impulse" -- you can also clap your hands or use an electric arc) and record the echo. This gives you what's called the "impulse response" of the space.

Reverb is really just thousands of echoes, and echoes are just the original sound delayed back to you. So what you can do is use the impulse response in a FIR filter, convolve it with the original signal, and you've recreated the same reverb with a different sound.

This technique has been quite accessible to audio engineers (and mechanical engineers) for a few decades now. But you really only go out and sample impulse responses when you really need a very accurate model of the reverb. In most cases someone will just use a precanned impulse response.

This is one area where hardware accelerated ray-tracing has me excited. I like the idea of walking sound sources to the observer. Unlike light, you need to account for the speed of sound through mediums, which adds an extra layer of cool imo. You also don't need to update at the audio sample rate, since most objects will not be moving very fast (unless you want trans sonic simulations, which would be exotic and cool).

Forget all of the little hacks. Sound processing is cheap. In 2020 it is unnecessary to select from a set of pre-baked filters and applying them to some predefined volume (ie this room is echoey, this room is absorbant, etc.)

This is one area where software has regressed then stagnated [0]. It doesn't make any sense, other than it wasn't prioritized. I think it's only a matter of time before engines go back and finally complete the audio simulation problem.

0. https://www.youtube.com/watch?v=cS_Fbueh7F4

I suppose it's a matter of acceptable approximations and cost. If only 2% of players have a high quality surround sound system, and only half of those have a good ear, then you really only have 1% of users who can tell the difference between raytraced audio and the heuristic "hacks" that sound engineers have spent decades developing. So I don't know if audio raytracing will become mainstream in games anytime soon. But I also suppose that if one engine does it, others will follow suit.

Same thing with this Unreal 5 demo. Most users would not be able to tell the difference between rendering 16B vs 8B triangles per frame.

You don't need a high quality sound system or a good ear to tell the difference though. You just need a pair of headphones and to be a human with hearing. Evolution has been at work for eons. Every human born is a powerful audio processing machine. I can't give you a rigorous proof or hand you a double blind test to try for yourself, so you'll have to accept an impasse or start going down a rabbit hole consisting of studies on pre-baked ray-traced FIR filters of known geometries and the audibility of group delay.

I am convinced every human can hear the difference. Computation is cheap. There's no reason not to solve the problem.

You should submit a patch to Godot, then! Force Unreal and others to implement it too.

This has been done for ages in the audio production space, it's called "convolution reverb". It's quite nice, but it doesn't allow you to tune it very much, besides basic filtering and stretching/cutting.

Yah, they said "convolution reverb," so I think they expected audience to know what it was. Often it's not about creating new technology, but making existing technology accessible to new audiences and new creators.

Honestly, convolution reverb is already highly accessible. It’s a feature in $50 (and probably free) reverb plug-ins for anybody that does audio. They’re relatively easy to implement and have probably been around for a decade+. As the OP said, the audio things they mentioned aren’t new or adding any accessibility imo

You have it right now in the browser Web Audio API, for free[1]

[1]: https://developer.mozilla.org/en-US/docs/Web/API/ConvolverNo...

They were doing it in the 80s for music when samplers first became big.

Yeah I figured so, but didn’t want to look up examples since I was on mobile and thus stuck with my more conservative value of decade+.

As a high-end builder, I would love to integrate a wall like this into a project. The ability (as shown with the Mandelorian video) the screens have to produce light would be incredible in interior living spaces.

Modular micro LED panels will make this easier, when they’re not absurdly expensive.

Reply on news.ycombinator.com