FYI the resulting product is <a href="https://www.newaudiotechnology.com/products/spatial-sound-card-l/" rel="nofollow noreferrer">https://www.newaudiotechnology.com/products/spatial-sound-ca...</a><p>(but I moved on to speech recognition a few years ago)<p>We improved the algorithm to allow for more customization and real time repositioning of the virtual speakers by mathematically decomposing the measured impulse data into components. Also, you need higher order approximations to get great bass, but the usual sources (e.g. my website and Wikipedia) only talk about IRs which are linear multiplications in frequency space.<p>But the most expensive part needed to turn it into a proper professional product was knowing people and paying for access to Hollywood monitoring studios. Because you want cinema sound, not a realistic approximation of your living room ;)
Interesting exercise, but for those looking for something closer finished product: look no further than openal-soft. Its HRTF support is what makes sound effects in Minecraft (Java Edition, on Windows and Linux -- headphone detection was very recently added for macOS) truly sound where they're coming from. Plus there's no intermediate mixing through any surround system, so you can even tell up from down. (HRTF in alsoft is not quite complete: it lacks support for near-field effects.)<p>Oh, (2014). That was a different time -- alsoft had just added HRTF and the technique was less known by then.
I've been going down this spatial audio rabbit hole for a few weeks now, and the tooling has left a lot to be desired. Admittedly, my own ignorance about tooling and audio mastering has probably been the largest barrier.<p>Ultimately I want a 3D environment where I can place a dummy head (the listener), and manipulate sound sources as objects while making them follow programmed trajectories. The best solution I've found so far seems to be Unreal Engine with some plugins, but progress quickly stalls as I'm overwhelmed by all of the technologies.<p>Honestly, considering how freaking cool this spatial audio can be, I'm amazed that simple creation tools aren't more widely available.<p>I enjoy listening to 3D ASMR recordings on YouTube but after a while most creators get a bit same-y. It feels like you could generate much more immersive spatial audio experiences, but only a small handful of channels are doing it.
Article is from 2014 but even in the days of AC97 codecs this sort of "3D sound" effect was already available in hardware. Realtek's drivers came with a utility to test this feature:<p><a href="https://www.youtube.com/watch?v=WZ0aHFwuCRo">https://www.youtube.com/watch?v=WZ0aHFwuCRo</a><p>Older versions used the more technically precise term "HRTF":<p><a href="https://infosys.beckhoff.com/content/1033/cx1020_cx1030_hw/Images/jpg/1489669387__en-US__Web.jpg" rel="nofollow noreferrer">https://infosys.beckhoff.com/content/1033/cx1020_cx1030_hw/I...</a>
Solutions have improved from a couple years ago, but in general, the issue I've found is that azimuth encoding wasn't working too well.<p>Maybe by interpolating on a lot of dynamically recorded impulse responses, that can be done better.