Zebralette 3.0 released (Revision 20399)

[Urs]

C & D are not used much because they were only just added.

[Rnaroth]

Audio Input is still on the table, but we need to do this later down the road.

Zebrify is not going away though. We're gonna keep it up to date for many years to come.

Zebrify is the most underrated plugin in the world. It is powerful in so many ways, there is nothing like it on the market. Glad it is getting the support it deserves.

[ashrwz]

where can i find the import setting like "torrance", "max points" etc., like in previous beta version?

[yp1]

Hey!
I was wondering about a potential feature:

How about a button like OSC source ([Curve Spectrum | Curve Geometry]), that instead changes the representation, not the interpretation of it?

So for example an L-shaped curve would be changed to a sine wave and back.
(Clicking that button and changing the OSC source would be a neutral operation then, at least theoretically/idealized.)

My current use case is as a learner of this curve-based synthesis -- getting a feel for how the drawing impacts the sound in the different interpretations sometimes brings up the question "huh, what would my curve look like in spectral/time domain?".

I'm somewhat aware that this might be tricky, and a solution would only ever be approximate (like going from pixels/wavetables to vectors generally is, I suppose).

Also, if this is already possible (without looking at a spectral analysis and then drawing manually or something), please educate me

[Andreya_Autumn]

Education incoming: If you want to see what it looks like, you can do that already by switching "Plot Domain" to frequency.

Automatically changing the curve to maintain the same sound when switching from Geometry to Spectrum interpretation is a really neat idea actually IMO. I am guessing it would be really hard to get the sound perfectly identical before/after the switch. Especially when going from spectrum to curve. A round trip would also naturally be lossy since the spectrum interpretation mode doesn't give any phase information. But that could be alright maybe.

[Urs]

Automatically changing the curve to maintain the same sound when switching from Geometry to Spectrum interpretation is a really neat idea actually IMO.

It is kind of coincidence. The falling sawtooth also has a similar spectrum. The spectrum is louder though, but that is totally natural and easily explained with math and geometry.

That is why we chose a falling sawtooth as a starting point, because a ramping one would sound terrible when switching between Curve Sources.

[Andreya_Autumn]

Terrible indeed... The spectrum/geometry switch is a dangerous one for a lot of curves. Any that end on a high value especially.

And yeah it is nicely peculiar how the shape of one saw cycle describes its own spectrum. I wonder if that's the only curve that does that. I think the answer might be yes but I couldn't prove it.

[yp1]

Education incoming: If you want to see what it looks like, you can do that already by switching "Plot Domain" to frequency.

Ah, thanks! I was using external tools before -- that's much better of course.

Automatically changing the curve to maintain the same sound when switching from Geometry to Spectrum interpretation is a really neat idea actually IMO. I am guessing it would be really hard to get the sound perfectly identical before/after the switch. Especially when going from spectrum to curve. A round trip would also naturally be lossy since the spectrum interpretation mode doesn't give any phase information. But that could be alright maybe.

Thanks, it's great to hear that it's not totally absurd

technically speaking (not that I understand this stuff, really -- just thinking out loud for the fun of it a bit):
- spectrum -> curve: "just" add a bunch of sines (?) -- no clue if this addition operation translates well to spline-world.. are the spline-based sines "perfect" in the first place there?
- curve -> spectrum: phew .. is that what Fourier transform does? I imagine at some point one has to decide on a resolution, and/or a way of fitting a curve to best-effort match a bunch of points
I suppose both could be a headache to implement, but I don't know much of course

musically speaking:
I love a lossy roundtrip!

[Andreya_Autumn]

I think you have spectrum and geometry reversed there.

You can indeed add individual sine waves with splines. In spectrum mode, set the grid to harmonics, then start with everything at 0 and put in whatever points you need to bring your desired harmonics up to their target amplitude. As you zoom in in the top range you get more resolution. A little tedious, but works well.

And the sines really are *extremely* precise. At least if you play a fundamental only, any noise/inaccuracies are ≈130-140dB or so below the scale of the sines themselves. That's about on par with the (also really precise) sines in Surge and Shortcircuit, which is kinda insanely precise for an additive renderer. Usually those compromise accuracy a bit to save CPU cycles since they need to make so many (and can still sound great). So yeah this is really impressive!

[Urs]

Our sine renderer uses a nifty trick to render near-perfect sines very fast. We favour sine accuracy and speed over phase accuracy. Earlier versions (Zebralette beta 1) would therefore drift a lot and quickly, but our newest version uses a small phase correction trick to prevent that drift, yet keep sines accurate and fast.

[yp1]

I think you have spectrum and geometry reversed there.

Yeah, I think I see what you mean.
I think should have written "spectral domain -> time domain".
"curve" (implemented as "splines") is really the representation layer, afaiu, so they're all curves, before and after the transform operation that we're discussing.

You can indeed add individual sine waves with splines. In spectrum mode, set the grid to harmonics, then start with everything at 0 and put in whatever points you need to bring your desired harmonics up to their target amplitude. As you zoom in in the top range you get more resolution. A little tedious, but works well.

Ok, after sleeping on it ...
The perfect-ish sines are definitely reachable through the "spectral domain" ("OSC source: Curve Spectrum") interpretation as you describe, good point!
In "time domain" ("OSC source: Curve Geometry") setting, it looks like (simple sums of) perfect sines are hard to get to ...
- the sine-like in the preset curve set produces some extra harmonics,
- I couldn't get rid of them -- and probably someone at u-he tried already
- (not saying the extra harmonics are actually audible, just looking at the visualization)
- at least from a naive standpoint that makes sense in that splines are basically polynomials (I suppose) and the sine function isn't

So, my current takeaway is that the two interpretations (spectral, time) aren't easily/naturally bridged on this level of curves.
That might be why no one has already made conversion switch I had in mind -- perhaps it seems messy/hacky to the experts .

I still feel curious, both about workflows being able to go back and forth between the two, as well as sounds that might come out. -- I can of course experiment by hand

Anyway, thanks for engaging with me on this!
Zebra3 is really interesting, on so many levels, and it's great that there's a place for these discussions

[exmatproton]

iwaswrong

[Future1313]

The Planet Moon Soundset by Sonic Sirius for the U-he - Zebralette 3 VSTi

- Inside the Download you get: 1 Planet Moon Soundset with 100 Presets
- Genres: Effects - New Wave - Modern Synth Music - Typical Zebralette
- Styles: 23x Effects, 10x Leads, 5x Pads, 4x Plucks, 58x Sequences
- For: U-he - Zebralette 3 v_300_20399 or newer
- Release: March 30, 2026
- Format: Presets (*.h2p)
- Size : 1,06 MB
- Price: Free

Install Instructions
Copy and Paste the " Sonic Sirius " Presets Folder into
C:\Users\YourName\Documents\u-he\Zebralette3.data\Presets\Zebralette3

Info: https://www.kvraudio.com/product/the-pl ... nic-sirius

Free Download
https://mega.nz/file/OqRnBKQY#XmJjn-biq ... HVS3ur3TrQ

[aMUSEd]

Thankyou

[3lu5iv3]

Thanks for the free preset pack! Future1313!