TL;DW: Dark is a wordlength reducer that gives your music a blacker backdrop.


Some weeks are MY kind of fun…

This won’t make big changes in your audio. In fact if you think you can reliably hear this on its HD setting, I think you’re mistaken. And yet, this might be the funnest thing I’ve done all year.

Why? Because I’m back to the dithers again. I figured out a way (or two) to go EVEN FARTHER in the direction I’d chosen. And it worked: it worked super well, and you can have it. Introducing Airwindows Dark.

How does it work? It’s very simple, really. Much like Not Just Another Dither (NJAD, my previous best) it analyzes the results of the audio, depending on whether the dither rounds up or down. With correct TPDF dither, it’s a factor of randomness, a noise that breaks up patterns in the output. With NJAD, it runs a Benford Realness calculation and uses that (for a more natural-sounding audio output). But Dark?

It simply works out the average trajectory of where the audio’s going. It’s following the lower frequencies, suppressing the highs. And then it makes its choice based solely on whatever is going to further this trajectory… and keep the output as smooth as possible. It is ‘dithering’ with intent, doing whatever it has to in order to get a darker, softer output. All done by truncation alone.

The result is delightful, if you are into analog sonics and the absence of bright digital artifacts and hisses. It is NOT obvious, unless 16 bit artifacts are already obvious to you, and at HD (24 bit) it is purely a matter of thoroughness and making correct choices and you shouldn’t be hearing a difference. You damn sure won’t be able to blind test a difference (that requires much more obvious stuff happening).

But, but, but! If your experience with audio (and probably loud listening levels, and REALLY good monitoring, and amazing source files) involves sinking into a lush bed of analog-rich sonics, where you quickly notice subtle shifts in sonic attitude and then take much longer to get used to them and form your judgements… there’s nothing at all like this.

It can only wordlength reduce, so especially at 24 bit it shouldn’t be able to ever hurt bright sounds that are supposed to be there. It’s only dithering (in a novel way). But it’s doing its thing in a way that’s completely outside of anything you can do with filtering or normal processing. If you need depth and space, if you need rich black silence behind your mix, this beats NJAD… soundly.

I hope you like it. The demonstration at 8 bit wordlength in the video ought to show you what to expect. Dark is yours to use (in fact, you can have the source code under the MIT license!). For now, if you are using Monitoring you’ll need to switch it off to use Dark, as Monitoring defaults to a 24 bit wordlength reduction using NJAD and I’m not prepared to simply update it and have it default to Dark for all things.

Though I’m tempted. ;)

This work is supported by Patreon. Which is going quite nicely, so I’m preparing for another phase: if I break $2000 a month, anything beyond that goes to buying DIY synth hardware (perfboard, CMOS chips, etc) which I will then resell at cost… MY cost. So I’ll be making chips available at 39 or 20 cents each and putting together kits to get people started, and each month I’ll send out stuff for people to play with, until I’ve reached the budget for that month. I may or may not charge shipping: haven’t decided. So if you think it would be good to start your own maker business and could use a cheap source of parts, the better I do the more likely you’ll be able to get your hands on electronics parts (and I will say where I’m getting stuff, if you need to order your own at normal prices: but I’ll be selling stuff at MY cost, no mark-up). I will also be writing up DIY guides and instructions, and doing videos and instagram posts about all this. That’s a new goal, because if I do better for myself I intend to spread it around in significant ways :)