Glad I’m not the only audio developer around here.
The landing page needs an immediate audio visual demo. Not an embedded YouTube but a videojs or similar. Low friction get the information of what it sounds and feels like immediately.
This looks incredible! But to be honest, it also looks incredibly daunting.
As a programmer and former physicist, I'm fascinated. As a musician, I'm not sure. At the moment, my feeling is that your landing page primarily addresses me as a programmer/physicist, and I'll definitely try it. But if you also want to sell this to musicians, what is really missing are more complex sound examples, like a tour of the existing presets and how you can manipulate them. There is your introduction video, but to be perfectly honest, the sounds you feature there do not really impress me. From what I can hear there, it very much sounds like the already existing physical modeling plugins, for instance AAS Chromaphone, and I already have plenty of those and they are much easier to use (also, their product page is a good example on how to sell a product to musicians). I can see of course that your VST allows me to dive much deeper into the weeds, and as a programmer/physicist I'm interested, but the musician in me is doubtful if the invested work will be worth with.
Again, this looks awesome, and I really hope you can make this into a business, so please see my critique above as encouragement.
Same here, and it is excellent.
I am getting a few buffer-drop clicking on an M3 MBP, reducing the polyphony solved it, but just in case, to the author: how much more efficiency you think you can still add to this amazing plugin?
This is a long story, which is still ongoing. The GPU code is very, very heavily optimized (though I do still have some ideas on how to go further). The main problem we're having on Mac hardware is that the OS heuristics for when to turn the clock rate up on the GPU work really poorly for the audio use case. If you want gory details, I've written about it:
For anyone seeing this post a bit late: I need a bit of help from someone inside Apple who works on Metal. If you know someone, it would be great if you could connect me to them:
I'm a huge fan of Animusic. I remember seeing it for the first time in some big fancy mall in LA and they had it projected on a wall, and I was blown away. It was absolutely an inspiration! Animusic -type ideas are a big part of why I made the 3D graphics fully user-customizable, for anyone who wants to go deep down that rabbit hole.
I'm several videos in and totally hooked, thank you for sharing. This would be an amazing interactive music app in VR, both to perform and to record trippy music videos.
I have the first two Animusic reels (vhs and dvd) and thought they were great. Unfortunately, the creator scammed people by taking money for Animusic 3 and then not making anything.
Not sure I'll ever use this as it seems like a lot of work but wanted to say thank you for allowing me to download a demo without giving an email.
Also, even though I said I wouldn't use it, something that would be nice is a master volume, maybe I missed it. I often use VSTs standalone and being able to change the volume without messing with the preset would make it a bit easier to use.
Definitely the most interesting synth I've ever seen.
Really cool stuff! I would suggest putting a 60-second video at the very top of the page that stitches together short clips of the many ways it is awesome.
Really cool to see GPUs applied to sound synthesis. Didn’t realize that all one needed to do to keep up with the audio thread was to batch computations at the size of the audio thread. I’m fascinated by the idea of doing the same kind of thing for continua in the manner of Stefan Bilbao: https://www.amazon.com/Numerical-Sound-Synthesis-Difference-...
Although I wonder if mathematically it’s the same thing …
Remind me of Korg's Berlin branch with their Phase8 instrument: https://korg.berlin/ . Life imitates art imitates life :)
I highly support and encourage this. Is there a way I could contribute to Anukari at all (I'm a physicist by day)? These kinds of advancements are the stuff I would live for! However I should stay rooted in what's possible or helpful: I'm not sure if this is open-source for example. As long as I could help, I'm game.
For the foreseeable future I'm just going to be working on stability/performance, but eventually I will get back to adding more cool physics stuff. It's not open-source, but certainly I'd enjoy talking to a real physicist (I'm something a couple notches below armchair-level). Hit me up at evan@anukari.com sometime if you like!
I was using the demo just now: the sounds you get out of this are actually better than I expected! And I see what you meant in the videos about intuitive editing, rather than abstract.
Although, I was often hitting 100% CPU with some presets, with the sound glitching accordingly. So I could experiment only in part. I'm on an M1 Pro; initially I set 128 buffer sample size in Ableton but most presets were glitching, I then set to 2048 just to check for improvement, which it did, nevertheless it does seem a bit high. Maybe my audio settings are incorrect? I can give more info later if it helps you.
Yeah performance at low buffer sizes is a big challenge, generally I recommend 512 or higher, which I know is not great but right now it's the most practical thing. The issue is that the computation is all done on the GPU, and there's a round-trip latency that has to be amortized. One day I'd like to convince Apple to work on the kernel scheduling latency...
I have been peripherally aware of Audiocube for a while, and it seems ridiculous that he and I have not interacted in any way. Maybe I'll bug him sometime. :)
At a glance, this looks like a bunch of coupled oscillators. A natural extension of this idea is strings: a 1d array of oscillators modelling a wave equation. For example, a piano sound can be modelled by attaching a basic oscillator to one end of a string and a mic to the other end of the string. The string and the oscillator push each other, creating the piano tone. Real pianos use 3 such string with different properties.
Another idea. What if you make a circular string and attach 1 or more oscillators at random points? Same idea as above, but more symmetric. This "sound ring" instrument may produce unreal sounds.
> What if you make a circular string and attach 1 or more oscillators at random points?
If your computer meets the system requirements, you could always install the free demo and build this sound ring instrument to find out! Building these kinds of weird ideas and seeing what happens is my favorite thing to do with it.
My RTX2080Ti (admittedly not a cheap card) supports 768 physics objects (masses) per each of 16 voices. And I think that beefier cards can do 1024. There are other limitations for performance depending on how the system is constructed, and I certainly don't want to claim that I know it can simulate a piano, but...
This is so cool and has unlimited potential, like you could model real instruments, e.g. guitar to experiment with resonant chamber shapes, materials etc. Can't upvote enough on good old perpetual licensing model!
I’m very curious about your experience doing audio on the GPU. What kind of worst-case latency are you able to get? Does it tend to be pretty deterministic or do you need to keep a lot of headroom for occasional latency spikes? Is the latency substantially different between integrated vs discrete GPUs?
Short answer: it has been a big pain in the butt. The GPU hardware is mostly really great, but the drivers/APIs were not designed for such a low-latency use case. There's (for audio) a large overhead latency in kernel execution scheduling. I've had to do a lot of fun optimization in terms of just reducing the runtime of the kernel itself, and a lot of less-fun evil dark magic optimization to e.g. trick macOS into raising the GPU clock speed.
Long answer: I've written a fair bit about this on my devlog. You might check out these tags:
Thanks for the extra info, I read through some of your entries on GPU optimization and it definitely seems like it's been a journey! Thanks for blazing the trail.
His stuff is so incredibly cool. He has a video on physical modeling for trumpets using the GPU and for a second I thought he might be building a competitor! :)
Note they are referring to Mick Gordon who is notable for the recent DOOM soundtrack. DOOM Eternal has a truly phenomenal score. Mick Cormick is a mistake I believe.
Oh goodness, that's truly embarrassing that I typoed "Mick Cormick" instead of Mick Gordon. O_o I wonder if my brain somehow crossed wires with John Carmack. Thanks for the correction!
Whoa this looks really cool! I love how you made something physically 3D to stand out in a world full of 2D knobs and sliders… but it still has 2D sliders because those work the best for dialing things in with precision.
I think to be useful it needs a mode of playing, that is always musical / in tune.
Not yet sure how to really do it, but one concept I like from NI plugins is that you have multiple keyboard zones: one zone is for notes, others are e.g. for patterns or styles. Imagine a guitar where one zone is for the chord type and tone, another for the striking pattern...
The challenge here is probably the resonance algo for multiple systems based on multiple notes... Maybe the piano concept would be handy here... imagine instead of having 3 strings like on the piano the instrument to be one system for each key... that excite each other via air or direct resonance points... the systems should be automatically tuned based on one reference system (e.g. using automatic string length or tension scaling)
Anyway, amazing work and having it on GPU allows this really to scale.
Absolutely awesome! I know nothing about music production but I want to play with it just for fun. Maybe a very simplified, web-based version for people who just to play a bit? Would be awesome.
Congrats on the hard work and the launch, in any case!
Edit: I see you have a demo mode, that's great! Exactly what I was looking for
I'm not really familiar with audio stuff, but holy do I ever appreciate the write-ups you've done. This is absolutely fascinating stuff. I'm eager to keep reading. The video from Mick Gordon was awesome, too.
Glad to hear it. I have fun writing them, often I find it's a great way to clarify my thoughts even just for me. But also I enjoy reading other people's devlogs so am glad to contribute. :)
My account shows 3 devices available to install on and I can disable computers on demand. Runs well on my M1 and on my 3060 and even all but the most demanding of assemblies on my little work laptop with onboard Intel graphics.
I assume files are compatible, presets are the same on both MacOS and Windows.
You are 100% correct. It turns out that I am an OK engineer, and a terrible marketer. There is a LOT that I need to improve on the site and the videos.
Every one of those pretty-looking screenshots should have a play button with a few seconds of audio. It's really strange to market a synthesiser with visuals instead of sound.
Yes, it's a synthesizer -- you may know it inside and out, but having demo videos showing what it can do will help people with no context get that quick "ahhh, that makes sense" moment from things. :)
I have mixed feelings about all of my supposed clever ideas being executed on by other people way ahead of me, but this is cool and you have my respect.
Haha, I know that feeling. Usually I get an idea, wait 4 years and someone else will have done it.
Even "bad" ideas. I had an idea in the mid 90s about mass internet surveillance. But then I thought it would be so disgusting that noone would do it. I was naive.
I like the perpetual license, no AI, and customisable 3D models and animations. This last feature hopefully opens up potential for making creatively synchronised graphics such as animating the expression - say the mouth shape on a 3D face in response to modulation. I wonder if the animation needs to be a fixed amount of frames or length?
The animations have some subtlety, but the basic idea is that they all to from t=0 to t=1, and then Anukari drives that t parameter. So for example, the mallet would go from resting at t=0 to fully extended at t=1, and Anukari will animate it at the right speed from 0, up to some value based on the velocity, back to zero.
Some of the objects are cyclic, like the spinning oscillators, and those also use the t=0 to t=1 convention, but require that the animation is seamless (so it looks identical at t=0 and t=1, but does whatever you want in between).
I'm not aware of anyone fully customizing the 3D stuff yet, so the instructions may be a bit rough. If you try it, feel free to contact me at evan@anukari.com or on discord and I'd be happy to answer questions.
Don't be discouraged! It might even be that 2D is better than 3D in this case: it's all about how it sounds, right? And if a 2D simulation can be less expensive than a 3D while sounding just as good or better, it works in your favour!
I think that's the real key to this stuff: what makes these things actually sound good?
There are a lot of advantages to 2D -- you could simulate more objects and more complex interactions with the lesser computational demands, and as other comments say, it will likely be way easier to build a better GUI. Think about how many compressor VSTs there are out there, and still people keep making them! And a 2D Anukari could be much more different from 3D Anukari than most compressors are from one another.
I think 2D is probably the better move for this thing. 3D doesn't really open many possibilities that you can't do in 2D, and it adds loads of UI awkwardness.
I am your target market, and I can't wait to try out the demo once I'm back from Superbooth and allowed to be distracted even a little.
... but I wanted to say that even with all of the glowing feedback, US$70 for a beta v1 soft synth is a big enough ticket that it will be off-putting to some and difficult to afford for others. Yes, there are many [much] more expensive virtual instruments, and this occupies a pretty unique space. But if you're open to feedback, this is my initial gut reaction.
One thing I am surprised by is that there's no mention of VR/AR ambitions. When I fantasize about 3D instruments, I do so in the context of wanting to interact with them in a space I inhabit. Does this speak to you as well?
Consider putting the purchase CTA inside the app, and just having a giant DOWNLOAD NOW button above the fold. The download for the demo and the real one should be the same download.
Then I don’t have to make a buying decision up front - I can get it on my computer and running first in all cases.
Looks very fun :-) Does anyone notice issues with the sound quality? In many of the examples I hear clicking: sometimes as if the attack is too high, or as if there is some kind of aliasing or sample rate issue, or just clipping. Probably noticeable with headphones. For instance in the announcement video at 3:11 or in the "J.S. Bach, Prelude in C Major (BWV 846)" video between 4.4s and 7.2s. It's somewhat visible if I load that audio in Audacity and turn on the spectrogram view with these settings: Logarithmic, 200 to 6000 Hz. Algorithm: Reassignment, 1024, Blackman-Harris, 1. Colors: 50, 40, 50.
What's odd is that I hear the glitches in in Firefox and in the file downloaded with yt-dlp, but not in Chromium. Is Google serving me bad audio on purpose?
Correction: some videos also do have glitchs on Chromium.
Screen-recording Anukari has been a bit of a challenge, as OBS works best while using GPU encoding, and also seems to do things that the GPU doesn't like in general (and Anukari uses the GPU). I suspect what you're hearing in the videos has to do with that. But also I'm sure that the model for the mic compression could be improved, and I'm not sure about the default attack time, etc.
If screen recording is actually the thing causing the issue you might try CPU encoding with one of the fast lossless codecs and doing the "real" encoding in a second pass later. As a bonus, software encoding should also give a higher quality result. That does require an SSD and quite a bit of free space though.
Have you tried HDMI output from the computer running Anukari to an HDMI capture dongle on a second system, so that all of the recording overhead is offloaded?
I'm very rusty at music, but I always had a big soft spot for unusual/unique synthesis methods. I'll be buying a copy as soon as I'm back at a desktop system :).
I really like doing deep optimization work, so probably the GPU stuff was more fun from an engineering geek perspective. But the physics was also super fun, experimenting with how to model each object, and so on. The physics is a little more playful, since changing a tiny bit of simple math often does fun/weird things.
RE strings/air, I have thought about it, but only a little! Down the road I really want to explore more physics objects. It's very fun to think about how I'd integrate strings into the world, especially w.r.t. how they'd interact and connect with the masses. It seems like there could possibly be some very cool ways to do it.
Hey evan
Just wondering, can you import 3d models into this envireonment? Im still pining for a less "code" driven environment for this than max/msp modalsys.
Awesome, say i import a bugle model, for arguments sake, does anakuri support a wind stream? I dont see air/wind listed as an object like you see bow and pluck
No wind-type model so far, though the bow model can do some rather flute-y sounds (in fact for the current bow model, I might even argue that it is better for pan flute stuff than sounding like a realistic bow).
Honestly I just haven't given wind models much thought yet! Long-term once it's stable and fast everywhere, I look forward to adding more physics features, and this is the kind of thing I'd love to look into.
The sounds seem to mostly be modulated sines with limited timbral variety? I'm not sure how https://youtu.be/NYX_eeNVIEU?t=179 got harmonic series out of the building blocks.
I plan to write a devlog post about this at some point, but at one time I wondered the same thing. It turns out that starting in 2D, when there is more than 1 spring, things get more complex than you expect: it turns non-linear. If the mass is moved exactly along the axis of the 2 springs in the demo you linked, then it's just a simple harmonic oscillator. But if it's moved OFF axis, the force from the springs is constantly changing direction. The force orthogonal to the spring's resting axis increases with the sine of the angle that the spring has been pulled off to. So along the off-axis it is not a harmonic oscillator.
If this doesn't make sense it's entirely my fault for explaining it poorly -- I really want to write this up properly with some diagrams because it's a surprising thing and quite interesting, at least to me.
(BTW that demo has other things going on like mic isotropy and analog oscillators attached to vibrate two of the masses, so my spring tangent is not the only thing producing harmonics.)
You can get a sense for the workflow in the quick-start tutorial: https://www.youtube.com/watch?v=Gva-BVuWbUA. It doesn't cover the advanced features but it should give you and idea what it's like!
Linux: I very much want to do this, but unfortunately it's lower priority than getting things rock-solid on windows/mac. It's not a gigantic amount of work, but it's not trivial. Hopefully I can do it once things calm down with the Beta.
CLAP: I'm using the JUCE framework for plugin integrations, which doesn't currently support CLAP. But their roadmap says that the next major version will support CLAP, and I will definitely implement that in Anukari. Not sure when JUCE 9 comes out though, it could be a while.
The thought has definitely occurred to me. It's always been in the back of my mind on the "if it shows some success..." list. Glad to hear there's interest, I think it would be really fun if I implemented proper multi-touch. There are some other details I'd need to think through, though, since right now it mostly assumes you have a MIDI keyboard, but on an iPad it's just begging for touch controls for a lot of that stuff.
Amazing, I do all my music on an iPad simply because if I open my laptop work will materialise. Now that I've put the idea in your head though, I think it's probably my duty to warn you that the market is a tiny fraction of the desktop market.
Yep! I have a lot of unit/integration tests that were a lot easier to write and more reliable by making the simulation fully-deterministic. It does produce slightly different results on different GPUs due to small differences in the FP operations. (For this application it's really beneficial to let the compiler go crazy with FP re-ordering to get speed.)
mutagen sums up my experience pretty well -- I have tested it on a laptop with a pretty wimpy Intel Iris chipset and it definitely works, but you might be limited in terms of how complex of a preset you can run. But there is a LOT of fun to be had with small presets so it may still be worthwhile. Be sure to install the absolute most up-to-date drivers from Intel directly, and use a larger buffer size.
I've been able to run it on an Intel laptop with integrated video. I haven't been able to test the most complex models / presets, I might give that a shot this weekend and see where it falls apart.
Sorry about that. It's just me working on it, and so far my personal tenet is that the product itself is the top priority, and all else including marketing comes second, so none of the website/youtube/etc are as good as I'd like. Possibly I'll soon have money to hire help with some of that, or I'll get the product to a place where I'm happy enough to work more on marketing myself.
I don't think Intel Macs have the grunt for this. Physical modelling of complex networks is pretty intensive.
You can take it a stage further and model networks of complex shapes like metal plates. That gets even more interesting because you get multiple resonant modes.
In the limit you could use finite element modelling to create precise simulations of acoustic instruments - like all of the strings in a piano, all of the dampers, the resonator, and the wooden enclosure.
But that's a brute force way to do it, and there isn't nearly enough compute available to make it happen in real time. (You might be able to do it on a supercomputer. I'm not aware of anyone's who's tried.)
Glad I’m not the only audio developer around here.
The landing page needs an immediate audio visual demo. Not an embedded YouTube but a videojs or similar. Low friction get the information of what it sounds and feels like immediately.
My 2 cents
Exactly. Had to scroll for ages to find anything to do with demo audio. A good demo song/track should be the first thing on the page, I think.
1000% - I had to be able to find something listenable
> Glad I’m not the only audio developer around here
There are a few of us :)
This synth is very cool. Highly original. Kudos.
This looks incredible! But to be honest, it also looks incredibly daunting.
As a programmer and former physicist, I'm fascinated. As a musician, I'm not sure. At the moment, my feeling is that your landing page primarily addresses me as a programmer/physicist, and I'll definitely try it. But if you also want to sell this to musicians, what is really missing are more complex sound examples, like a tour of the existing presets and how you can manipulate them. There is your introduction video, but to be perfectly honest, the sounds you feature there do not really impress me. From what I can hear there, it very much sounds like the already existing physical modeling plugins, for instance AAS Chromaphone, and I already have plenty of those and they are much easier to use (also, their product page is a good example on how to sell a product to musicians). I can see of course that your VST allows me to dive much deeper into the weeds, and as a programmer/physicist I'm interested, but the musician in me is doubtful if the invested work will be worth with.
Again, this looks awesome, and I really hope you can make this into a business, so please see my critique above as encouragement.
OK, I've played around with the demo and insta-bought it, if just to support you. This is incredible work.
Same here, and it is excellent. I am getting a few buffer-drop clicking on an M3 MBP, reducing the polyphony solved it, but just in case, to the author: how much more efficiency you think you can still add to this amazing plugin?
This is a long story, which is still ongoing. The GPU code is very, very heavily optimized (though I do still have some ideas on how to go further). The main problem we're having on Mac hardware is that the OS heuristics for when to turn the clock rate up on the GPU work really poorly for the audio use case. If you want gory details, I've written about it:
https://anukari.com/blog/devlog/waste-makes-haste
If anyone can put me in touch directly with an OS/Metal person at Apple it would be EXTREMELY helpful. I've had limited success so far.
For anyone seeing this post a bit late: I need a bit of help from someone inside Apple who works on Metal. If you know someone, it would be great if you could connect me to them:
https://news.ycombinator.com/item?id=43901619
https://anukari.com/blog/devlog/an-appeal-to-apple
This reminds me of the reverse, where music drives 3D animations. I remember Animusic from the early decade of 2000.
https://en.wikipedia.org/wiki/Animusic , https://www.animusic.com/ , https://www.youtube.com/results?search_query=animusic , https://www.youtube.com/@julianlachniet9036/videos
I'm a huge fan of Animusic. I remember seeing it for the first time in some big fancy mall in LA and they had it projected on a wall, and I was blown away. It was absolutely an inspiration! Animusic -type ideas are a big part of why I made the 3D graphics fully user-customizable, for anyone who wants to go deep down that rabbit hole.
This rings such a vague and distant bell...
I'm several videos in and totally hooked, thank you for sharing. This would be an amazing interactive music app in VR, both to perform and to record trippy music videos.
I have the first two Animusic reels (vhs and dvd) and thought they were great. Unfortunately, the creator scammed people by taking money for Animusic 3 and then not making anything.
Most of them are on youtube.
Not sure I'll ever use this as it seems like a lot of work but wanted to say thank you for allowing me to download a demo without giving an email.
Also, even though I said I wouldn't use it, something that would be nice is a master volume, maybe I missed it. I often use VSTs standalone and being able to change the volume without messing with the preset would make it a bit easier to use.
Definitely the most interesting synth I've ever seen.
Thanks, yeah, it really should have master volume -- you didn't miss it, it's just not there yet!
Really cool stuff! I would suggest putting a 60-second video at the very top of the page that stitches together short clips of the many ways it is awesome.
Really cool to see GPUs applied to sound synthesis. Didn’t realize that all one needed to do to keep up with the audio thread was to batch computations at the size of the audio thread. I’m fascinated by the idea of doing the same kind of thing for continua in the manner of Stefan Bilbao: https://www.amazon.com/Numerical-Sound-Synthesis-Difference-...
Although I wonder if mathematically it’s the same thing …
Thank you for this, it looks very cool!
Remind me of Korg's Berlin branch with their Phase8 instrument: https://korg.berlin/ . Life imitates art imitates life :)
I highly support and encourage this. Is there a way I could contribute to Anukari at all (I'm a physicist by day)? These kinds of advancements are the stuff I would live for! However I should stay rooted in what's possible or helpful: I'm not sure if this is open-source for example. As long as I could help, I'm game.
For the foreseeable future I'm just going to be working on stability/performance, but eventually I will get back to adding more cool physics stuff. It's not open-source, but certainly I'd enjoy talking to a real physicist (I'm something a couple notches below armchair-level). Hit me up at evan@anukari.com sometime if you like!
Thanks, will hit you up later!
I was using the demo just now: the sounds you get out of this are actually better than I expected! And I see what you meant in the videos about intuitive editing, rather than abstract.
Although, I was often hitting 100% CPU with some presets, with the sound glitching accordingly. So I could experiment only in part. I'm on an M1 Pro; initially I set 128 buffer sample size in Ableton but most presets were glitching, I then set to 2048 just to check for improvement, which it did, nevertheless it does seem a bit high. Maybe my audio settings are incorrect? I can give more info later if it helps you.
Yeah performance at low buffer sizes is a big challenge, generally I recommend 512 or higher, which I know is not great but right now it's the most practical thing. The issue is that the computation is all done on the GPU, and there's a round-trip latency that has to be amortized. One day I'd like to convince Apple to work on the kernel scheduling latency...
I would love to watch (and listen) to a discussion between you and Noah from Audiocube, https://news.ycombinator.com/item?id=42877399 https://main.audiocube.app/ a 3d spatial DAW.
I have been peripherally aware of Audiocube for a while, and it seems ridiculous that he and I have not interacted in any way. Maybe I'll bug him sometime. :)
At a glance, this looks like a bunch of coupled oscillators. A natural extension of this idea is strings: a 1d array of oscillators modelling a wave equation. For example, a piano sound can be modelled by attaching a basic oscillator to one end of a string and a mic to the other end of the string. The string and the oscillator push each other, creating the piano tone. Real pianos use 3 such string with different properties.
Another idea. What if you make a circular string and attach 1 or more oscillators at random points? Same idea as above, but more symmetric. This "sound ring" instrument may produce unreal sounds.
> What if you make a circular string and attach 1 or more oscillators at random points?
If your computer meets the system requirements, you could always install the free demo and build this sound ring instrument to find out! Building these kinds of weird ideas and seeing what happens is my favorite thing to do with it.
A piano string would have to be made of about 1000 basic oscillators.
My RTX2080Ti (admittedly not a cheap card) supports 768 physics objects (masses) per each of 16 voices. And I think that beefier cards can do 1024. There are other limitations for performance depending on how the system is constructed, and I certainly don't want to claim that I know it can simulate a piano, but...
This is so cool and has unlimited potential, like you could model real instruments, e.g. guitar to experiment with resonant chamber shapes, materials etc. Can't upvote enough on good old perpetual licensing model!
I’m very curious about your experience doing audio on the GPU. What kind of worst-case latency are you able to get? Does it tend to be pretty deterministic or do you need to keep a lot of headroom for occasional latency spikes? Is the latency substantially different between integrated vs discrete GPUs?
Short answer: it has been a big pain in the butt. The GPU hardware is mostly really great, but the drivers/APIs were not designed for such a low-latency use case. There's (for audio) a large overhead latency in kernel execution scheduling. I've had to do a lot of fun optimization in terms of just reducing the runtime of the kernel itself, and a lot of less-fun evil dark magic optimization to e.g. trick macOS into raising the GPU clock speed.
Long answer: I've written a fair bit about this on my devlog. You might check out these tags:
https://anukari.com/blog/devlog/tags/gpu https://anukari.com/blog/devlog/tags/optimization
Thanks for the extra info, I read through some of your entries on GPU optimization and it definitely seems like it's been a journey! Thanks for blazing the trail.
Love physics based audio! Using the GPU is a great idea.
Another physical audio simulation I like is the engine sound simulator made by AngeTheGreat: https://youtu.be/RKT-sKtR970?si=t193nZwh-jaSctQM
His stuff is so incredibly cool. He has a video on physical modeling for trumpets using the GPU and for a second I thought he might be building a competitor! :)
Note they are referring to Mick Gordon who is notable for the recent DOOM soundtrack. DOOM Eternal has a truly phenomenal score. Mick Cormick is a mistake I believe.
Congratulations!!
Oh goodness, that's truly embarrassing that I typoed "Mick Cormick" instead of Mick Gordon. O_o I wonder if my brain somehow crossed wires with John Carmack. Thanks for the correction!
McCormick is a popular brand of seasonings hahaha
https://i5.walmartimages.com/seo/McCormick-Pure-Ground-Black...
Of note is the drama surrounding the Doom Eternal soundtrack: https://medium.com/@mickgordon/my-full-statement-regarding-d...
So many of us have ideas for something cool and never build them. You did it. I salute you, you madman
Whoa this looks really cool! I love how you made something physically 3D to stand out in a world full of 2D knobs and sliders… but it still has 2D sliders because those work the best for dialing things in with precision.
I think to be useful it needs a mode of playing, that is always musical / in tune.
Not yet sure how to really do it, but one concept I like from NI plugins is that you have multiple keyboard zones: one zone is for notes, others are e.g. for patterns or styles. Imagine a guitar where one zone is for the chord type and tone, another for the striking pattern...
The challenge here is probably the resonance algo for multiple systems based on multiple notes... Maybe the piano concept would be handy here... imagine instead of having 3 strings like on the piano the instrument to be one system for each key... that excite each other via air or direct resonance points... the systems should be automatically tuned based on one reference system (e.g. using automatic string length or tension scaling)
Anyway, amazing work and having it on GPU allows this really to scale.
Absolutely awesome! I know nothing about music production but I want to play with it just for fun. Maybe a very simplified, web-based version for people who just to play a bit? Would be awesome.
Congrats on the hard work and the launch, in any case!
Edit: I see you have a demo mode, that's great! Exactly what I was looking for
I'm not really familiar with audio stuff, but holy do I ever appreciate the write-ups you've done. This is absolutely fascinating stuff. I'm eager to keep reading. The video from Mick Gordon was awesome, too.
Congratulations on the launch, and best of luck!
Glad to hear it. I have fun writing them, often I find it's a great way to clarify my thoughts even just for me. But also I enjoy reading other people's devlogs so am glad to contribute. :)
I’m so tempted to buy, but some info is missing on the website:
- If I buy once can I run it on both my Windows desktop and MacBook travel computer?
- If so, are files compatible between them?
- What are GPU requirements on Windows? I’m sure it scales, but is a 3080 overkill or not enough?
My account shows 3 devices available to install on and I can disable computers on demand. Runs well on my M1 and on my 3060 and even all but the most demanding of assemblies on my little work laptop with onboard Intel graphics.
I assume files are compatible, presets are the same on both MacOS and Windows.
Yep! I can confirm everything mutagen said. Presets are portable between mac and windows for sure.
Thank you!
The website should have a better youtube video and at the beginning
You are 100% correct. It turns out that I am an OK engineer, and a terrible marketer. There is a LOT that I need to improve on the site and the videos.
Every one of those pretty-looking screenshots should have a play button with a few seconds of audio. It's really strange to market a synthesiser with visuals instead of sound.
Yes, it's a synthesizer -- you may know it inside and out, but having demo videos showing what it can do will help people with no context get that quick "ahhh, that makes sense" moment from things. :)
Some little audio examples would also be nice so visitors don't have to scroll through the video to hear them.
Still, awesome work!
Yeah, all the cool 3D pictures should play demo videos on click!
>a terrible marketer
I wouldn't go so far, apart from this point the landing page is excellent.
I have mixed feelings about all of my supposed clever ideas being executed on by other people way ahead of me, but this is cool and you have my respect.
Haha, I know that feeling. Usually I get an idea, wait 4 years and someone else will have done it.
Even "bad" ideas. I had an idea in the mid 90s about mass internet surveillance. But then I thought it would be so disgusting that noone would do it. I was naive.
This is insane. I've used tons of virtual synths in my life but this is by far the coolest I've ever seen. Mick's video was amazing!
I like the perpetual license, no AI, and customisable 3D models and animations. This last feature hopefully opens up potential for making creatively synchronised graphics such as animating the expression - say the mouth shape on a 3D face in response to modulation. I wonder if the animation needs to be a fixed amount of frames or length?
Some info about custom 3D models is here: https://anukari.com/support/faq#custom-skins
The animations have some subtlety, but the basic idea is that they all to from t=0 to t=1, and then Anukari drives that t parameter. So for example, the mallet would go from resting at t=0 to fully extended at t=1, and Anukari will animate it at the right speed from 0, up to some value based on the velocity, back to zero.
Some of the objects are cyclic, like the spinning oscillators, and those also use the t=0 to t=1 convention, but require that the animation is seamless (so it looks identical at t=0 and t=1, but does whatever you want in between).
I'm not aware of anyone fully customizing the 3D stuff yet, so the instructions may be a bit rough. If you try it, feel free to contact me at evan@anukari.com or on discord and I'd be happy to answer questions.
Dang it. I am working on the same thing, but in 2D.
Don't be discouraged! It might even be that 2D is better than 3D in this case: it's all about how it sounds, right? And if a 2D simulation can be less expensive than a 3D while sounding just as good or better, it works in your favour!
I think that's the real key to this stuff: what makes these things actually sound good?
There are a lot of advantages to 2D -- you could simulate more objects and more complex interactions with the lesser computational demands, and as other comments say, it will likely be way easier to build a better GUI. Think about how many compressor VSTs there are out there, and still people keep making them! And a 2D Anukari could be much more different from 3D Anukari than most compressors are from one another.
I think 2D is probably the better move for this thing. 3D doesn't really open many possibilities that you can't do in 2D, and it adds loads of UI awkwardness.
3d is cooler, but 2d is easier to use for people, there was some research on it, though don't have it at hand
edit: a hn thread https://news.ycombinator.com/item?id=19961812
I am your target market, and I can't wait to try out the demo once I'm back from Superbooth and allowed to be distracted even a little.
... but I wanted to say that even with all of the glowing feedback, US$70 for a beta v1 soft synth is a big enough ticket that it will be off-putting to some and difficult to afford for others. Yes, there are many [much] more expensive virtual instruments, and this occupies a pretty unique space. But if you're open to feedback, this is my initial gut reaction.
One thing I am surprised by is that there's no mention of VR/AR ambitions. When I fantasize about 3D instruments, I do so in the context of wanting to interact with them in a space I inhabit. Does this speak to you as well?
Ha! I've had this idea for ages but never had the urge to make it. I'll have to play with it, thanks for sharing!
Consider putting the purchase CTA inside the app, and just having a giant DOWNLOAD NOW button above the fold. The download for the demo and the real one should be the same download.
Then I don’t have to make a buying decision up front - I can get it on my computer and running first in all cases.
Looks very fun :-) Does anyone notice issues with the sound quality? In many of the examples I hear clicking: sometimes as if the attack is too high, or as if there is some kind of aliasing or sample rate issue, or just clipping. Probably noticeable with headphones. For instance in the announcement video at 3:11 or in the "J.S. Bach, Prelude in C Major (BWV 846)" video between 4.4s and 7.2s. It's somewhat visible if I load that audio in Audacity and turn on the spectrogram view with these settings: Logarithmic, 200 to 6000 Hz. Algorithm: Reassignment, 1024, Blackman-Harris, 1. Colors: 50, 40, 50.
What's odd is that I hear the glitches in in Firefox and in the file downloaded with yt-dlp, but not in Chromium. Is Google serving me bad audio on purpose?
Correction: some videos also do have glitchs on Chromium.
Screen-recording Anukari has been a bit of a challenge, as OBS works best while using GPU encoding, and also seems to do things that the GPU doesn't like in general (and Anukari uses the GPU). I suspect what you're hearing in the videos has to do with that. But also I'm sure that the model for the mic compression could be improved, and I'm not sure about the default attack time, etc.
If screen recording is actually the thing causing the issue you might try CPU encoding with one of the fast lossless codecs and doing the "real" encoding in a second pass later. As a bonus, software encoding should also give a higher quality result. That does require an SSD and quite a bit of free space though.
Have you tried HDMI output from the computer running Anukari to an HDMI capture dongle on a second system, so that all of the recording overhead is offloaded?
I'm very rusty at music, but I always had a big soft spot for unusual/unique synthesis methods. I'll be buying a copy as soon as I'm back at a desktop system :).
So beautiful - I wonder what kind of an instrument an AI can build, creating sounds never before heard...
Sounds like a lot of work has gone into to the infrastructure of getting this to work on the GPU.
Did you find it more interesting doing this or the physics simulation ?
Have you considered simulating strings or moving air too ?
I really like doing deep optimization work, so probably the GPU stuff was more fun from an engineering geek perspective. But the physics was also super fun, experimenting with how to model each object, and so on. The physics is a little more playful, since changing a tiny bit of simple math often does fun/weird things.
RE strings/air, I have thought about it, but only a little! Down the road I really want to explore more physics objects. It's very fun to think about how I'd integrate strings into the world, especially w.r.t. how they'd interact and connect with the masses. It seems like there could possibly be some very cool ways to do it.
Seriously cool! A Linux build on the horizon?
https://news.ycombinator.com/item?id=43873635
I second the recommendation: even if you don't have a Twitter/X account, find some way to watch Mick Gordon's session!
I find it hysterically funny, but at the same time, it really shows what this synth is capable of.
Excellent!
Hey evan Just wondering, can you import 3d models into this envireonment? Im still pining for a less "code" driven environment for this than max/msp modalsys.
You can replace all the 3D models and skyboxes. Everything is in open formats, wrapped up in zip files. See:
https://anukari.com/support/faq#custom-skyboxes https://anukari.com/support/faq#custom-skins
AFAIK nobody has attempted this yet, so the write-up might not be perfect. If you try it, let me know how it goes!
Awesome, say i import a bugle model, for arguments sake, does anakuri support a wind stream? I dont see air/wind listed as an object like you see bow and pluck
No wind-type model so far, though the bow model can do some rather flute-y sounds (in fact for the current bow model, I might even argue that it is better for pan flute stuff than sounding like a realistic bow).
Thanks Evan, is that because wind models are complex to model?
Honestly I just haven't given wind models much thought yet! Long-term once it's stable and fast everywhere, I look forward to adding more physics features, and this is the kind of thing I'd love to look into.
Your beta video assumes I know why I should care and jumps straight into technicals.
This is potentially new to producers. Tell them why they should care first.
This thing is insane, got me wanting to mess around with it even though I can barely play keys.
The sounds seem to mostly be modulated sines with limited timbral variety? I'm not sure how https://youtu.be/NYX_eeNVIEU?t=179 got harmonic series out of the building blocks.
I plan to write a devlog post about this at some point, but at one time I wondered the same thing. It turns out that starting in 2D, when there is more than 1 spring, things get more complex than you expect: it turns non-linear. If the mass is moved exactly along the axis of the 2 springs in the demo you linked, then it's just a simple harmonic oscillator. But if it's moved OFF axis, the force from the springs is constantly changing direction. The force orthogonal to the spring's resting axis increases with the sine of the angle that the spring has been pulled off to. So along the off-axis it is not a harmonic oscillator.
If this doesn't make sense it's entirely my fault for explaining it poorly -- I really want to write this up properly with some diagrams because it's a surprising thing and quite interesting, at least to me.
(BTW that demo has other things going on like mic isotropy and analog oscillators attached to vibrate two of the masses, so my spring tangent is not the only thing producing harmonics.)
Very nice to see. Maybe nice in VR, streaming to Quest 3/Apple Vision Pro (OpenVR/WebXR?)
I don't have the first idea about audio but damn this looks powerful.
Man. “Sheet music is not portable”. I immediately thought about the Apple Vision Pro
The battery-less medium is something we do desperately try to mimic
It definitely looks really cool! As an outsider to the audio stuff with an okay amount of knowledge I’m curious as to the workflow for sure
You can get a sense for the workflow in the quick-start tutorial: https://www.youtube.com/watch?v=Gva-BVuWbUA. It doesn't cover the advanced features but it should give you and idea what it's like!
Any chance we'll get a Linux VST or CLAP?
Linux: I very much want to do this, but unfortunately it's lower priority than getting things rock-solid on windows/mac. It's not a gigantic amount of work, but it's not trivial. Hopefully I can do it once things calm down with the Beta.
CLAP: I'm using the JUCE framework for plugin integrations, which doesn't currently support CLAP. But their roadmap says that the next major version will support CLAP, and I will definitely implement that in Anukari. Not sure when JUCE 9 comes out though, it could be a while.
There's an external plugin to emit a CLAP plugin for JUCE 8: https://github.com/free-audio/clap-juce-extensions
It works quite well, but it's also reasonable to wait for official framework support.
Thanks, that could be good if the JUCE support is going to be way off.
Hopefully they make it easy. WINE has some issues with JUCE 8 I think.
It seems like vst problems with WINE always comes down to issues with license auth and graphics libs.
Incredible work and a very creative product. I can't wait to see what is created using Anukari.
This is crazy, incredible work.
Wow. This is so cool. It opens up a different approach to the problem.
I like the look of this, do you have any plans to release it for iPad?
The thought has definitely occurred to me. It's always been in the back of my mind on the "if it shows some success..." list. Glad to hear there's interest, I think it would be really fun if I implemented proper multi-touch. There are some other details I'd need to think through, though, since right now it mostly assumes you have a MIDI keyboard, but on an iPad it's just begging for touch controls for a lot of that stuff.
Amazing, I do all my music on an iPad simply because if I open my laptop work will materialise. Now that I've put the idea in your head though, I think it's probably my duty to warn you that the market is a tiny fraction of the desktop market.
I'd focus on ipad & an awesome multitouch experience. App store sales are easier & i'd bet apple would spotlight it.
Very neat!
Is the simulation deterministic?
Yep! I have a lot of unit/integration tests that were a lot easier to write and more reliable by making the simulation fully-deterministic. It does produce slightly different results on different GPUs due to small differences in the FP operations. (For this application it's really beneficial to let the compiler go crazy with FP re-ordering to get speed.)
does it benefit from or even require top end GPUs to get the best results?
mutagen sums up my experience pretty well -- I have tested it on a laptop with a pretty wimpy Intel Iris chipset and it definitely works, but you might be limited in terms of how complex of a preset you can run. But there is a LOT of fun to be had with small presets so it may still be worthwhile. Be sure to install the absolute most up-to-date drivers from Intel directly, and use a larger buffer size.
I've been able to run it on an Intel laptop with integrated video. I haven't been able to test the most complex models / presets, I might give that a shot this weekend and see where it falls apart.
Cool idea. Sounds like AAS Chromaphone on steroids.
very impressive. is it built with juce?
Yep!
is this about sound generation? Because I didn't find any sample sounds on this wall of text and pictures.
Sorry about that. It's just me working on it, and so far my personal tenet is that the product itself is the top priority, and all else including marketing comes second, so none of the website/youtube/etc are as good as I'd like. Possibly I'll soon have money to hire help with some of that, or I'll get the product to a place where I'm happy enough to work more on marketing myself.
Ah yes the new The Incredible Machine sequel is looking good.
This is super super duper cool. Thx for sharing
Fun :)
Not supported: Intel-based Macs
Boo
I don't think Intel Macs have the grunt for this. Physical modelling of complex networks is pretty intensive.
You can take it a stage further and model networks of complex shapes like metal plates. That gets even more interesting because you get multiple resonant modes.
In the limit you could use finite element modelling to create precise simulations of acoustic instruments - like all of the strings in a piano, all of the dampers, the resonator, and the wooden enclosure.
But that's a brute force way to do it, and there isn't nearly enough compute available to make it happen in real time. (You might be able to do it on a supercomputer. I'm not aware of anyone's who's tried.)
I feel like you're overlooking the fact that it's GPU based (sorta the whle point) and that's why it also runs on Windows.
Whoa this is sick
looks really cool! congrats!
This is so cool.