Zebra 3 and Zebra Legacy

[nilhartman]

I for one can't wait to have additive oscillators in Z3. *Crossing fingers* some of the partials manipulations one can do with Razor, Lazerbass, Harmor and/or, indeed, Loom 2, will be doable with Z3.

The kind of super sharp sounds you can hear here :
https://gaborlazar.bandcamp.com/track/boundary-object-v
https://kindohm.bandcamp.com/track/prompt-2
https://errorsmithmarkfell.bandcamp.com ... -atomic-80

Urs, I was wondering if Z3 oscillators will be able to be an FM carrier and to use its additive rendering engine at the same time ?

[Urs]

Urs, I was wondering if Z3 oscillators will be able to be an FM carrier and to use its additive rendering engine at the same time ?

This is an interesting idea that I haven't thought about yet.

The renderer is highly complex and extremely optimised - it is as of yet one of two DSP algorithms in Z3 that we have optimised for AVX2 and AVX512 in addition to our common SSE/NEON code. If there is a chance to somehow FM up to 2048 sine oscillators, I'll surely give it a go. It would not be DX-style FM though, as the sine oscillators are not generated from phase information. I think it's rather be exponential or linear FM, and probably promote inharmonic sounds rather than harmonic FM.

[Urs]

The kind of super sharp sounds you can hear here :

Yeah, those, I reckon, quite easily.

[nilhartman]

Marvelous, thanks Urs !

[GruvSyco]

I’ve been obsessed lately with the Nonlinear Labs C15. I’ve been trying to replicate the chain as best I can in Zebra. The way it does feedback doesn’t appear to be doable but it’s sent me down a rabbit hole of exploration I’ve only dabbled with in Zebra. While I was scraping the forum looking for tips and tricks, I found a discussion on Zebra FM (Yamaha style) and how there was one algorithm that couldn’t be replicated because the feedback loop spans two oscillators. I kinda hate being one of those “Urs can you do this” people but, it could be cool to have definable feedback loops or maybe a send and return within the feedback loop.

It’s cool that I’m still discovering things in Zebra.

[Urs]

There's actually two algorithms I think. One spans two oscs, one spans three.

Unfortunately there's no good way to bend Zebra's concept into something that allows freely routable feedback between modules. The routing is always column by column, row by row in Zebra's Grid.

I had explored sending delayed signals from one Grid cell to one arbitrary other in Zebra 1.5, but it wasn't really all that great. Not sure if trying this again would yield better results unless we drastically increase CPU load.

[GruvSyco]

unless we drastically increase CPU load.

Isn't that what you do though? Break defenseless CPUs and make them cry.

CPUs always seem to catch up. DIVA and Bazille are nowhere near as crushing on my M1 as they were on whatever x86 I was running when they came out.

[Urs]

Yeah, hehehe, I kind of had my share of customer grief... I try to be not as bad for now...

[robisme]

[quote=GruvSyco post_id=8790888
[/quote]
Isn't that what you do though? Break defenseless CPUs and make them cry.
[/quote]

He he, place a mic inside the computer and catch the sound of the burning. This is then routed to a zebra xmf module.

[DrWashington]

Very much agree about Loom II on both points: sounds superb, the UI is an ugly, dated mess that doesn't appear compatible with reasonably modern computers. I really, really wish they'd pick it back up. I miss my K5000S and I've not come across any additive synth that's a viable replacement. It was a PITA to program with any level of granularity, but it could certainly make sounds nothing else could. It filled a unique niche, and I miss that.

I would love it if someone made an additive synth with the goal of 100% transparency for resynthesis and maybe used AI to examine what would most likely be the most usable macros for humans and then present, say, 16 of them to bind to MIDI controllers right away. But of course you'd still have complete control of each envelope of each partial with as many breakpoints as you want, and the envelopes could be fast enough and randomizable such that it could even resynthesize noise elements transparently.

I think additive is a super untapped field, and I think with the advent of AI, it could be made far more useful. In fact, deep learning could even be employed to "understand" what it is your resynthesizing and decide on a user-available parameter palette based on its best guesses.

If 100% transparency were the goal first and foremost, this could be a truly transformational hypothetical product.

[bonch]

I’ve been obsessed lately with the Nonlinear Labs C15. I’ve been trying to replicate the chain as best I can in Zebra. The way it does feedback doesn’t appear to be doable but it’s sent me down a rabbit hole of exploration I’ve only dabbled with in Zebra. While I was scraping the forum looking for tips and tricks, I found a discussion on Zebra FM (Yamaha style) and how there was one algorithm that couldn’t be replicated because the feedback loop spans two oscillators. I kinda hate being one of those “Urs can you do this” people but, it could be cool to have definable feedback loops or maybe a send and return within the feedback loop.

A user-definable feedback loop is one thing that keeps me going back to Massive X for certain effects.

[Dirtgrain]

I had explored sending delayed signals from one Grid cell to one arbitrary other in Zebra 1.5, but it wasn't really all that great. Not sure if trying this again would yield better results unless we drastically increase CPU load.

These are the things that can keep one up at night.

[GruvSyco]

Unfortunately there's no good way to bend Zebra's concept into something that allows freely routable feedback between modules. The routing is always column by column, row by row in Zebra's Grid.

So, I’m sitting on the couch watching Indiana Jones and the Last Crusade and an idea pops into my head:

Feedback module is a forced pair (linked… both on or off) or one module with two locations. Both can pass through audio. Feedback 1 (in) has a level (from Feedback 1 (out)) and a mod source (because everything needs modulation). Feedback 1 (out) is basically just two identical signals, the pass thru and the feedback loop. Overly simplified, with no idea of coding.

Anyways… I won’t go on on this suggestion. There are better minds at work here than mine.

[GruvSyco]

A user-definable feedback loop is one thing that keeps me going back to Massive X for certain effects.

This whole idea (on my part) started in pursuit of replicating something that already exists as another NI product. Kontour is a direct descendant of the Nonlinear Labs C15.

It must be taxing to a developer having everyone suggesting inclusion of everything. At least I’m not asking for sampling.

[dayjob]

Very much agree about Loom II on both points: sounds superb, the UI is an ugly, dated mess that doesn't appear compatible with reasonably modern computers. I really, really wish they'd pick it back up. I miss my K5000S and I've not come across any additive synth that's a viable replacement. It was a PITA to program with any level of granularity, but it could certainly make sounds nothing else could. It filled a unique niche, and I miss that.

I would love it if someone made an additive synth with the goal of 100% transparency for resynthesis and maybe used AI to examine what would most likely be the most usable macros for humans and then present, say, 16 of them to bind to MIDI controllers right away. But of course you'd still have complete control of each envelope of each partial with as many breakpoints as you want, and the envelopes could be fast enough and randomizable such that it could even resynthesize noise elements transparently.

I think additive is a super untapped field, and I think with the advent of AI, it could be made far more useful. In fact, deep learning could even be employed to "understand" what it is your resynthesizing and decide on a user-available parameter palette based on its best guesses.

If 100% transparency were the goal first and foremost, this could be a truly transformational hypothetical product.

SUMU is due soon. Maybe it will check some of those boxes