Between samples and synth, sound design and performance—and with a heaping dose of machine learning—Visco is a new way of imagining a beat construction plug-in. But what you’ll probably notice first is that you get to squish around a futuristic blob of goo like it’s space-alien Silly Putty.

The music instrument software and hardware business is always full of people heading off on their own to try new ideas. And that’s what’s happened with Svante Stadler and Rikard Jönsson, who recently founded the creative studio Forever 89 after stints at some of those bigger makers. Visco, launched today, is their first product. And it’s pretty wild.

The usual boilerplate applies – new ways to make sound more accessible, opening up sound design possibilities without the interface… okay, okay.

Let’s get straight down to it.

One, you can squish around goo-like shapes and make extreme sound mutations.

Two, yes, readers experienced with “AI” will get that there’s some machine learning at work. Specifically, think machine learning and the ability to explore latent space – read this explanation if you want to know more about the topic. The basic idea is, you can pull in any sample, derive a number of characteristics from it, and then morph between different sounds and discover “latent” possible mutations of your sounds.

Three, despite the AI and 3D interface elements, you still get a bunch of precise controls when you want them, including the parameters controlling variations and modifications, more conventional velocity and voicing parameters, plus modulation, including four assignable envelopes and four LFOs.

That is, this is not actually a dumbed-down “gooey AI drum kit for n00bs” – it has all the advanced features you might want, too. But you know what advanced sound designers will be doing? Well, squishing and experimenting with extremes, too, of course!

The usual machine learning artifacts apply in the form of noise and characteristic harmonic content. But that’s honestly part of the appeal here, as you can get some pleasing metallic and pitched and dirty sounds you wouldn’t get via other methods. And with a little more precision, you can also make some dynamic, morphing sounds that are more realistic, too, if you so desire. They’ve done some detailed processing work on the engine, so it honestly sounds pretty excellent – not grating in the way some other AI-powered sonic engines can be. You do start to get the sense that this can be another approach to synthesis, not just a gimmick.

They must feel the same, as you don’t once read “AI” or “machine learning” anywhere in the site or press materials. It seems the developers are resisting the hype. And unlike generative AI, you’re exploring sounds directly with local data, not just typing in a text prompt that squishes together data via massive amounts of processing.

It can sound really terrific, as you can hear in the bonus tracks – and the “flaws” are part of that character, meaning unless you’ve recently put a ton of work into DIYing your own latent space drum sampler plug-in, you won’t find anything that sounds quite like this:

In that central display, you get a unique cursor that lets you squish around a graphic representation of the sound, morphing multiple parameters at once into different shapes like putty. There are also graphical tools for taking those parameters and mirroring and rotating them – the kind of graphical tools we’ve seen far too little in music outside of (now U&I Software, originally MetaCreations) MetaSynth.

You also have color-coded drum parts, and both the sequencer and main view show each sound not by name but by waveform envelope.

Inside, there are a bunch of features:

  • 32-voice, 8-part drum engine
  • Per-part and global controls for transform, timescale, frequency, contrast (brings components in and out of the sound), and density (noise component)
  • 8-track sequencer with swing and time/velocity variations, per-step velocity, and the ability to drag note events and lengths (including off the grid, so it’s a piano roll and step sequencer in one)
  • Modulation page with assignable grid (source, target, scalar) – tons of options there
  • Mixer with per-track filter, clipping distortion, pan, stereo width, two sends
  • Send and master effects, including chorus, diffusers, reverbs (plate, room, spring), tempo delay, stereo fuzz, various compressors (including upward compressor), tape simulator, and transformer amp

In short, they really plussed the heck out of this thing, and it seems they’re just getting started. (See the screens here.)

The sample collection alone is already fun to navigate, and you can import your own:

The sequencer works internally, and even with a free-running clock (or sync to your host). But you can also easily trigger via MIDI – or, going the other way, export MIDI from Visco back to your host or other sequencer.

I’d love to see dedicated parameter lanes in the sequencer. For now, though, you do get polymeters – just drag that last space:

There are also a surprising number of parameters you can route or assign. All the parameters can also be modulated externally, so of course, I tried some external Live LFOs as well as the internal ones.

You can directly edit envelope and LFO shapes:

The mixer offers plenty of extras, as well:

Video exploration

While making absolutely no claim of musical quality here, le’s give this a try! Here’s a live jam (cut for time), building up a weird morphing beat from scratch, then adding some external modulation (just because you can) and effects. If you encourage me, I might do a more proper sound demo of the internal engine and effects; let me know if that’s of interest.


Visco is available now for macOS 10.13 and later (Apple Silicon/Intel) AU and ST3, and Windows 10+ AU/VST3.

It’s currently on an intro price of €99 before going up to 139. There’s a free unlimited demo. Copy protection is textfile-based.

Available direct from Forever 89 (not to be confused with Cycling ’74 or Forever 21).

The goos I’ve known before

All of this will remind some of you of Beatsurfing’s recent, wonderful RANDOM (which I’ll also review ASAP – even shot a video already) and the ongoing otherworldly goodness that is Synplant 2. But while those also use related approaches and UI ideas, they sound very different. If this is a trend, I am there for it. Sheesh, the last thing we need is another subtractive synth or vintage recreation of the same gear or simple sample-playback-based instrument. I for one welcome our new alien invaders.

And that means it’s also a good time to remember Kai’s Power Goo, which really nailed the interface elements of “take something onscreen and squish it around.” Plus the story behind it and Kai Krause remain forever compelling.

Oh, and if you want some research links on related audio and machine learning, here’s a place to start:

An Exploration of the Latent Space of a Convolutional Variational Autoencoder for the Generation of Musical Instrument Tones

But who has time for that? I’m going to go look for an emulator to run Kai’s Power Goo on and then soundtrack it with Visco. Obviously.