Anukari Synthesizer: Sound through the laws of physics

[runagate]

I certainly will, I always watch every video from the ADC anyway. Also, I stopped paying attention to your progress for the time being because I get so ludicrously hyped-up wanting to start sound designing.

[anukari-music]

Long time, no update! I was busy for some time preparing for my ADC talk (which went really well), but now after the holidays I'm back to full time engineering work on Anukari.

Things were slow, but that's not to say that nothing got done. When I had time, I was working on all the machinery under the hood to allow modulators to be added: LFOs, envelopes, MIDI control, DAW params, etc. Today I finally got this to the point where there's something small I can demo:

https://www.youtube.com/watch?v=ybzamZJr83k

The modulation framework is now mostly in place, and there's just a bunch of work to add all the different kinds of modulators, as well as making every parameter I can a modulation target. I am really hoping to be able to make modulators themselves modulation targets (e.g. use an LFO to modulate another LFO's frequency, etc). But the GPU is not making that particularly easy to do. Still, I think I can probably do one layer of LFO recursion, but not more. So no LFOs modulating LFOs modulating LFOs.

[runagate]

You might want to talk to the GPU audio folks.
Also, I need this software.

[anukari-music]

> You might want to talk to the GPU audio folks.

I got to talk to them at ADC! Alexander Talashov specifically. Very nice guy and we had a good discussion of what they're doing and what capabilities they offer. My problem space is a bit weird as it's not traditional DSP processing, but rather is this weird physics simulation, but still it is possible that GPU Audio could be useful for the cross-platform part. I will look into it when I'm happy with the features and want to get it up and running fully on Mac.

I suspect that I will still just port the physics to Apple's MTL shader language myself, but from talking to Alexander, it does seem that in principle it would be possible to use GPU audio for the portability. But I'm a bit iffy on the code parsing / generation they do to make that possible. And more to the point, I'm just a bare metal kind of person, and am always a bit skeptical about extra layers of abstraction. But for sure I'll play around with their offering.

[neverbefore]

damn, this looks very interesting but also looks like it is just for mac

[runagate]

I suspect that I will still just port the physics to Apple's MTL shader language myself, but from talking to Alexander, it does seem that in principle it would be possible to use GPU audio for the portability. But I'm a bit iffy on the code parsing / generation they do to make that possible. And more to the point, I'm just a bare metal kind of person, and am always a bit skeptical about extra layers of abstraction. But for sure I'll play around with their offering.

Well, that is certainly illuminating. Now, I am no coder but it immediately makes me think that having the physics, visual models and audio directly coupled sounds incredibly daunting. What we end users rarely realize is that not only is sound generation a hell of a lot less costly than the other processes but that many decades of incredibly optimized algorithms are being used. Which makes me think how odd it is our industry has not adopted some sort of audio card standard with FPGs which can be set up as algorith-specific DSP like the XMOS chip.

[anukari-music]

damn, this looks very interesting but also looks like it is just for mac

Actually at the moment it only runs on Windows. It probably would run on Linux since it's using OpenCL, but I haven't tried that. I 100% intend to make it run on Mac later, and have designed it to be "easy" to port, but as usual the devil is in the details. The problem is that Apple deprecated OpenCL (which they created, lol), so it will require more work to support their new MTL language.

[neverbefore]

damn, this looks very interesting but also looks like it is just for mac

Actually at the moment it only runs on Windows. It probably would run on Linux since it's using OpenCL, but I haven't tried that. I 100% intend to make it run on Mac later, and have designed it to be "easy" to port, but as usual the devil is in the details. The problem is that Apple deprecated OpenCL (which they created, lol), so it will require more work to support their new MTL language.

hooray - so far seems you are making a wonderful new synth that I will almost certainly buy (if I can afford it):)

[anukari-music]

Well, that is certainly illuminating. Now, I am no coder but it immediately makes me think that having the physics, visual models and audio directly coupled sounds incredibly daunting. What we end users rarely realize is that not only is sound generation a hell of a lot less costly than the other processes but that many decades of incredibly optimized algorithms are being used. Which makes me think how odd it is our industry has not adopted some sort of audio card standard with FPGs which can be set up as algorith-specific DSP like the XMOS chip.

It is indeed daunting! But all solvable, I think. When I first got started on Anukari, I took the approach of starting with the hardest problems -- the ones I was not at all certain that I could solve. That way, I'd find out I couldn't do it nice and early and move on to something else. But all the things I thought I might not be able to figure out how to do turned out to be tractable. At this point I am fairly certain there's nothing unsolvable left, just a big pile of work.

I agree that it's odd that a powerful audio card DSP standard has not really happened. Especially with Apple trying to position Macs as the go-to for music production, you'd think they'd have crammed in some ultra DSP stuff on their M2 chips.

I think to some extent GPUs may fill the role of standard DSP in the future, since they are ubiquitous at this point. But they are... weird. You only get their power if you can (massively) parallelize your algorithm. That's often possible, but sometimes doing so is really difficult.

[anukari-music]

This morning I published a long-form walk-through demo where I go through all the different features in Anukari. I go through all the kinds of physics objects you can create and how they work, and I also show the basic features that the user interface has and how you interact with it. After I've shown how everything works, I use it to create a guitar effect and play some guitar through it, to give folks a sense of what the workflow is like.

Still tons of work to go, but hopefully this video gives folks a sense for where Anukari is at today and where it might be going!

https://www.youtube.com/watch?v=iYJcVXjRi-I

[tonycore]

Looking great! This is going to be a truly unique addition to the arsenal...

[runagate]

<insert take my money meme here>

[tbugz]

I never imagined the hours spent playing "Worms" in my youth would actually be useful in my music making many years later, but there you have it!
In all seriousness though, hats off to you for an amazing effort

[anukari-music]

Thanks, everyone!

Here's another demo video where I build a super-simple polyphonic instrument from scratch, since the last demo covered an effects processor but not MIDI:

https://www.youtube.com/watch?v=zr7sHmZr6Vs

[runagate]

I love this