Hi All.
I was just wondering what you think about my recent inquiry here. Are they fair questions and answers for example and/or would you add any of your own? What else?
____

Spoiler: ai

Q: How many sound engines does VirSyn CUBE 2.2 have at 512 partials per voice?

AI: VirSyn CUBE 2.2 features 8 independent additive synthesizer engines, each capable of up to 512 partials per voice, enabling complex, evolving sounds through spectral morphing and detailed parameter control.

Q: So if each engine has 512 partials per voice and we can use all 8 engines, does this mean we can get 512 × 8 — 4,096 partials — in total when hitting a key on our keyboard/MIDI controller? In other words, can each engine operate at the same time with every other engine?

AI: Yes, each of the 8 engines in VirSyn CUBE 2.2 can operate simultaneously, and each supports up to 512 partials per voice. When you trigger a note, all 8 engines can be active at once, meaning the synthesizer is capable of producing up to 4,096 partials in total (8 × 512) across all engines. The effective number per voice depends on how the engines are configured, but all 8 can contribute to a single voice.

[...]

The architecture uses four morphable sound sources (A, B, C, D), and each source can run two engines, totaling 8. These can be combined and modulated together, allowing all 8 engines to contribute to one voice when programmed accordingly.

Q: How would or might this compare with the upcoming Zebra 3 synthesizer by U-He? If you are unsure, please look at U-He's Zebralette 3 for insight, since it is supposed to be a simple version of Zebra 3.

AI: U-He Zebra 3 features a new additive/wavetable oscillator engine derived from Zebralette 3, supporting up to 1024 partials in additive mode—more than CUBE 2.2’s 512 per engine. However, Zebra 3 does not have 8 fixed additive engines. Instead, it uses a modular architecture with flexible oscillator routing. While it supports fewer parallel additive engines, it offers greater spectral control, spline-based editing, and hybrid synthesis capabilities. Zebra 3 prioritizes depth and versatility over sheer engine count.

Q: CUBE 2.2 also offers resynthesis. In this or a similar regard, how might Zebra 3 compare?

AI: VirSyn CUBE 2.2 supports resynthesis by analyzing audio and converting it into additive spectra with up to 512 partials, enabling morphing between modeled sounds. Zebra 3 does not offer direct resynthesis from audio files. Its additive oscillator allows manual harmonic shaping using spline curves, but it cannot automatically analyze and reconstruct external audio. Therefore, CUBE 2.2 is superior in automated resynthesis, while Zebra 3 focuses on advanced manual spectral design.
____

Q: Given that we can import and resynthesize a sound in Cube 2.2 and morph 4 different sounds (or maybe 2 sounds per morphed sound), how fast, quick and convenient is it in theory to generate useful sounds in it compared with Zebra 3, like with less manual fiddling and whatnot?

AI: VirSyn CUBE 2.2 is specifically designed for fast, intuitive sound generation through resynthesis and morphing. It allows direct import of WAV/AIFF files, automatically analyzing them into additive spectra with up to 512 partials. You can load and morph between four resynthesized sounds using macro controls, enabling dramatic, evolving timbres with minimal manual editing. The interface includes "fruit machine" randomization and macro parameters that make discovering usable, complex sounds quick and efficient—ideal for experimental and cinematic sound design.

In contrast, Zebra 3 does not support automated resynthesis. While it offers powerful manual spectral shaping (up to 1024 partials) and flexible modulation, creating similar results requires more hands-on, parameter-by-parameter adjustment. Its strength lies in precision and hybrid design, not rapid resynthesis-based workflows.

Thus, CUBE 2.2 is significantly faster and more convenient for generating rich, evolving sounds from imported audio with minimal fiddling, while Zebra 3 demands more manual effort for comparable complexity.

Q: With regard to VirSyn Cube's subsequent and relatively recent desktop 64-bit Cube (Cube 2 Pro or something like that), is our dialogue still relevant to Cube 2.2? If so, how and, if not, how? In other words, any new or deprecated features?

AI: Yes, our dialogue remains relevant. The desktop version of VirSyn CUBE was updated to 64-bit in 2014 under the same CUBE 2.2 name, with no major architectural changes. There is no separate product called "Cube 2 Pro"—the core engine, 8 additive synthesizer engines, 512 partials per engine, and resynthesis/morphing capabilities are unchanged. The 64-bit update ensured compatibility with modern systems but did not add new synthesis features. Development focus has since shifted to iOS apps (e.g., Cube Synth Pro for iPad), but the desktop CUBE 2.2 remains functionally identical to the version discussed.

Q: How many additive oscillators (presumably at 512 partials per voice or oscillator?) will Zebra 3 be able to run at the same time, such as when pressing a key on a MIDI keyboard?

AI: Zebra 3 does not use traditional additive oscillators with fixed partial counts like VirSyn CUBE 2.2. Instead, it features a single, highly advanced oscillator per voice derived from Zebralette 3, capable of up to 1024 partials in additive/wavetable mode. This oscillator supports spline-based spectral shaping, morphing, and effects—but it is not multiplied across multiple instances. While Zebra 3 allows flexible modulation and layering via its modular architecture, each voice uses one primary oscillator engine that can blend or switch between synthesis types, rather than running several discrete additive oscillators in parallel. Therefore, when pressing a key, Zebra 3 uses one powerful, multifunctional oscillator per voice, not multiple additive oscillators like CUBE 2.2’s 8-engine setup.

Q: If Zebra 3's oscillator engine can, as you write, blend or switch between multiple synthesis types, how many multiple synthesis types can be blended or switched at the press of a key? Can the blending be between 2 additive ('wavetable') oscillators at 1024 partials each? If so, how many additive oscillators can be used for this blending or switching?

AI: Zebra 3 uses one advanced oscillator per voice, which can generate additive, wavetable, VA, or modal sounds—but only one type at a time per oscillator. It does not support multiple simultaneous additive oscillators. This single oscillator can morph between different spectral shapes or synthesis types via modulation (e.g., using XY pads or envelopes), but it cannot layer two independent additive sources at 1024 partials each. Therefore, while blending between synthesis characteristics is possible, Zebra 3 does not allow parallel additive oscillators like VirSyn CUBE 2.2.

 

Last edited by a moderator: Nov 27, 2025 at 12:04 PM

Your report reads like a marketing advertisement from VirSyn - Cube against its competitor Zebra 3.
Despite the facts, U-he - Zebra 3 sounds significantly better than VirSyn - Cube.

 

What do you mean by 'better' and have you heard Zebra 3? Is there an alpha or beta out or are you referrring to Zebralette?

Zebra 3 appears more of a 'kitchen-sink' synth too, whereas Cube 2 is purely additive.

As an aside, Sumu's got a problem with my computer's memory.

 

Last edited: Nov 27, 2025 at 12:02 PM

According to Urs at Kvraudio the public beta is in the works after the Black Friday sale ends:

Hard to make any concrete comparisons or even specwise ones before then. but I do find it hard to believe that Z3 will have only one oscillator like Zebralette3 as Z2 has four Zebralette2 oscillators.

 

Last edited: Nov 27, 2025 at 1:13 PM

To me, Zebra definitely sounds better, more powerful, more precise—one of the best-sounding virtual synthesizers. That's thanks to Urs, who builds high-quality synthesizers; he has the best code, and that's why Zebra was and is one of the most widely used and highly praised virtual synthesizers.

Additive synthesizers usually sound very thin and weak, but you can still love the sound. A good-sounding additive synth is Air Music - Loom, and the recently released discoDSP - Vertigo 2: https://www.soundonsound.com/reviews/disco-dsp-vertigo-2.0

Cube 2 never really caught on in the market, meaning it's not widely used.
Sylenth 1, Ana, Spire, and Nexus all sound better than Cube 2.

What is additive synthesis?
Unlike subtractive synthesis, additive synthesis doesn't subtract anything from the sound; instead, it adds to it. Usually, two or more waveforms of the same or different frequencies are superimposed. Unless you're working exclusively with pure sine waveforms (we'll get to that in the next episode), certain harmonics are canceled out (so not just added after all ;)) and others are added during these superpositions. This makes it very easy to create new waveforms and dynamically modify them. Unless you're working with pure sine waves (we'll get to that in the next episode), certain harmonics are canceled out (so not just added after all ;)) and others are added.

I created a soundset for the Tera 3 myself:

VirSyn - Tera 3 VSTi
- The 'Northern Lights' Soundset by CHE for the VirSyn - Tera 3 VSTi
https://ohlenbostelhelge.wixsite.com/sonic-sirius-sound-d
Free Download: https://mega.nz/file/jvoSVBpS#usD-qnFdcvsxk_cLHPcAfSKbyS8MlpIYHLwcRrzqUyg

Are you getting an error message? Please provide some more information. What CPU, etc., do you have?

 

Good to know and agreed.

I wonder if additive is more of an afterthought in 'kitchen-sink'-style synths. I suppose we'll find out soon enough in Z3's case.

 

Zebra is a wireless modular synthesizer that combines subtractive and additive synthesis with a powerful modulation engine and built-in effects section. ---> www.kvraudio.com/product/zebra-by-u-he
VirSyn Cube is a Spectral morphing additive synthesizer.

Additive synthesis never really caught on; you're more likely to find FM (frequency modulation) and granular synthesis. In better synthesizers, we also find other filters like Comb + Comb and Karplus. Strong string synthesis is a method of physical modeling synthesis that loops a short waveform through a filtered delay line to simulate the sound of a hammered or plucked string or some types of percussion.

 

You're referring to Z2, right? Is it additive? In what sense?

I've always suspected that Urs was somehow 'cheating' and that his synths were somehow all granular and/or granular-wavetable at heart. Urs even wrote something once over at KVR that seemed to underscore my suspicions about it. Ditto with Myth, BTW.
I still have my doubts that granular synths-- or some or all wavetable synths for that matter-- are 'true' or 'pure' synths. People have of course called wavetable synths plasticky and/or somehow static.

Sometimes synths-- like anything-- like Cube perhaps-- can arrive at the wrong time and place, but, yes, 'better' seems in the ears of the beholder.

Cube and/or additive may have 'caught on', just not necessarily in a mass-adoption sense. Maybe in a cult following.

Then again, I remind that two additive synths-- Harmor and Alchemy-- were enclosed in DAWs. So maybe they didn't 'catch on' exactly, because they were 'stolen'.

Still, though, I seem to like to approach my instruments as instruments beyond just the sounds they make and I made such an argument over on Vital's site-- the idea that some of us use instruments not just because of the sound they make but because of how they make the sounds or how we 'feel' they do, as well as how they look and feel in our hands and so forth.

That's ok about Sumu, incidentally. I haven't pursued the issue just yet, and/since Madrona Labs (Sumu's dev) are ostensibly working on an update, including or especially where memory is concerned. It is known as being a bit memory-hogging.

I also normally do my own sounds, but thanks just the same. If I change my mind, I might give you a shout.

My Linplug Spectral sounds 'thick' and 'strong'-- even 'better' (so to speak) than my OsTIrus.

Loom 2 sounds good too, but it's problematic for at least a couple of reasons.

 

Last edited: Nov 27, 2025 at 1:20 PM

LinPlug also invented and used the Golden Code, which makes LinPlug sound simply better than average. Zebra is always Zebra; it's just improved. Check out "Reason - Europa," it's amazing: https://www.musicradar.com/reviews/propellerhead-europa

You shouldn't always make the same mistake and directly compare synthesizers; they all have their purpose and function.

 

Golden Choice

I may have added more to my comment above that you just replied to, so you may see fit to re-check it out.

But wow, the Golden Code, what's that? Still, though, I'm unsure I should be surprised. So Linplug takes off and doesn't opensource their Golden Code and/or at least Spectral?

I have Europa in my Reason Rack (RR), thanks-- sort of matryoshkaed in a DAW/Rack that I don't normally use. I got RR for Parsec. Too many synths, too little time. ;P

I think it's fine to compare anything. That's, for example, part of the process by which we, or some of us, choose our musical instruments. Or much of anything else for that matter. Why do you prefer apples over oranges or vice-versa?

'Do you prefer the piano? Why? Why not? The flute? The vibraphone? Why?'

Stuff like that.

BTW, of course if we use stuff that doesn't catch on, maybe that gives us a more unique sound and separates us more from the crowd as it were.

Ever heard of Uhhyou? I should do a project using just their synths.

 

Last edited: Nov 27, 2025 at 1:51 PM

LinPlug closed its doors years ago. https://www-amazona-de.translate.go...l=auto&_x_tr_tl=en&_x_tr_hl=de&_x_tr_pto=wapp

The "Golden Code" refers to the programming; he was a very good coder, with extensive expertise and a love for detail that made Linplug one of the best plugins.

 

Sure, why not? For one, it can be part of the creative process, yes? To create, do you always do stuff that's 'expected'?

There's always something that we can get out of particular excercises, too, that may not necessarily be obvious or make much sense... say, such as bringing new perspectives on something that's about to come out that we may not have considered before. Like before we hand over our hard-earned money. Or the idea of the myth of progress. Stuff like that.

At the very least, call it a pre-emptive comparison if you wish. I won't mind. :D

Maybe 'old school' isn't so old, too, I mean, given the (over-?)abundance of emulations and all that.

 

Last edited: Nov 27, 2025 at 2:08 PM

Old-School Software Nostalgia Yet?

Incidentally, I think Cube 2 has been updated, at least a little, but even so, apparently it was a bit ahead of its time anyway. That's in part what inspired me with this thread. I caught you talking about it on the sister site. I forget if it was you, specifically, who said to the effect that it was 'ahead of its time', but you seemed to appreciate it too.

And so, here I make a thread 'comparing' a synth that is about to arrive with one that was ahead of its time. <-- There is something to that, maybe something important I don't know.

Edit: Almost forgot to mention that I also caught something last week of an increase in an interest now in old-school music/audio software, along with this video below with what looks to be dated UI's, etc..

(BTW, thanks to you or whomever for the spoiler thing. Makes things neat and tidy.)

 

Last edited: Nov 27, 2025 at 2:29 PM

What an odd choice to compare those two, it's like an apple and a carrot. To what purpose that would be? To see if they can morph additive synths sources? If so, they both can, but differently.
Zebra is a (semi)modular powerhouse rightfully revered as on off the best synths of all times because how versatile it can be while Cube is an experimental additive resynthesis. Usable sounds in Cube2 is a happy accident, you get what you get and then you are more or less stuck with that with Zebra, however, you'll be able to sculpt it further with "indefinite" precision.

What that suppose to mean? How do you quantize "pureness"??? Especially considering that the functionality of a WTsynth is in the name itself, it scans wavetables rather than compute a predefined set of OSC shapes, but it doesn't make it "dirty" or any less "true" than VA synths. It's not like WT synths use samples like PCM synths (Korg M1, Roland D-50 etc) to generate sounds

 

Ya, but you've got a carrot in the apple, that's the thing. That's even more of an odd comparison, but compare we must if one is going to stick/kludge a carrot in an apple. Serves them right, yes? :D

Apparently, granular synths, maybe wavetable too, or some of them, use the audio signal as-is so to speak. They are not turned into pure sine partials. That's what I mean. Granular is just chopped up audio and wavetable may just be more of the same, just 'samples' if understood correctly. There doesn't appear to be any hardcore resynthesis into partials and working with them, and nothing else.

So you might as well get into cutting and splicing analogue tape or at least be honest with yourself. (Just joking but only half so.)

Do you want a real synthesizer or a glorified audio-tape-cutter?

So what we seem to get are 'kludges' as kitchen-sink synths and some 'cheating' thrown in. Just a whole lot of whole lot thrown in. Ok, so Pigments does it prettily with flowing rainbow waves. But how well does it do additive, or is it just an afterthought, compared with pure additives?

Unsure what you mean by usable sounds, but I'll have to cut my teeth on Cube 2 to see.

 

Last edited: Nov 27, 2025 at 2:46 PM

1. LinPlug Spectral was programmed in C++.

LinPlug plugins, including Spectral, are high-performance audio plugins (VST/AU) optimized for real-time synthesis. C++ is the standard language for such applications because it enables low latency, efficient DSP processing (e.g., additive synthesis with 256 harmonics and 121-band filters), and cross-platform compatibility (Windows/Mac).

2. u-he Zebra (and Zebra2, ZebraHZ, Zebralette) was programmed in C++, just like almost all u-he plugins.

Confirmed details:
Urs Allemann ("Urs," the founder and lead developer of u-he) programs all his synthesizers himself in pure native C++ (without frameworks like JUCE in the early years).

Zebra2 uses a highly optimized, hand-written DSP engine in C++ with intensive use of SSE intrinsics (later also AVX) to enable extremely complex, real-time modular synthesis.

For a long time, the GUI was implemented with custom-written C++ GUI tools (no JUCE, no VSTGUI). It wasn't until around 2018/2019 (e.g., with Diva Revision 2, Hive 2, Repro) that u-he gradually began replacing its own GUI-based

 

Best thread for wasting time: where AI and pointless discussions collide… and debating while sanity sneaks out for a cigarette

 

I'm speaking about what's under the hood so to speak, rather than the code used to program the synth. For example...

Spoiler: synths under the hood

"A wavetable synth oscillator works by reproducing a waveform that has been recorded digitally—in other words, converted into a string of numbers corresponding to the amplitude of the waveform in a sequence of moments. By reading through this list of numbers repeatedly, the oscillator reproduces the recorded waveform...

The recorded waveform could be anything we can capture (or create) digitally. We could record a snippet of a beloved old analog synth, or an acoustic instrument. We could even capture a bird call, or the sound of the wind.

This is just sampling, right? Not exactly. A sampler typically reproduces longer sounds—perhaps of a few seconds, which may feature many different waveforms over their duration. By contrast, a wavetable synth captures a single cycle of a waveform and reproduces it as the basis for a sound. This is called sample-based synthesis."
____

"Additive synthesis is infinitely powerful. It is limited only by the technical limitations of your sound-generation machinery, and by your time, skill, understanding, and imagination. That said, it is also the most-demanding, upon all of those resources. You can literally synthesize any sound that has ever existed, or that ever could exist, by adding together some sequence and combination of pure sine waves. The downside is that sound-design tends to be a technical, laborious, analytical, and somewhat un-musical process, without an easy or intuitive way to "tune" or dial in sounds on-the-fly. Plus, you typically need much more powerful and expensive hardware to produce sounds that are often no better than other synthesis methods. Still, it is theoretically the ideal system."

 

:D <-- old-school smiley
<3
...So I sit down to a taste-test. There are two ice creams. I don't know which is which, but find that I prefer, at least taste-wise, the one that has the most artificial crap in it. I don't even know what half the ingredients are or are for. But from other standpoints, I prefer the one with the natural ingredients.

So which do I choose?

The one with the natural ingredients. It might not taste as good, but it feels better.