Madrona Labs Sumu

[WasteLand]

It takes 85% of my CPU

all cores? at 8 voices steady, almost at 64% about.... on 1 core by the way, which is no problem.


(liked you bandcamp link by the way, really liked it! forgot to mention it!!!)

It takes 85% of my CPU

It's not quite that bad for me, but it's the only synth I use where I can't get away with a buffer size of 32 samples (unless I limit it to being monophonic). 96 works if I'm careful about what else is running. It's a hungry, hungry hippo...

never did that, setting my sound interface to 128 (i don't have 96, 64 i do have, in ms..).
haha.
well it works as good as in 256 ms, even 64 ms will work...

i never see much improvement when upping the buffer size, in my situation, a Ryzen 9 5900X with RME Multiface (bridged PCI...). o well i don not always try it.. but i does not matter that much.
5900x i believe gets lazy at higher buffer size, and that has a mind of its own, how to set the clock speed of one core (of 12/24 cores). i don't use a hard overclock (and even in that case), the PBO overclock, with undervolting per core.

to much hardware info.... but it sometimes not very predictable it seems... or predictable it is, but you must consider some factors, a always.

"γνῶθι σαὐτόν" ~ kingtubby

And your surrounds.

o well, it is KVR, but quoting, or citing that Delphi thingy seems very very. o well. quoting Herakleitos, and to know what your are citing.. and to know what 'citing' means! well that is step forward. and don't use the wrong citation of Herakleitos and i know Herakleitos has nothing to with; "γνῶθι σαὐτόν"

'Reverse Engineering' Sumu?

Could Sumu be somehow duplicated, echoed or improved in Plugdata? Probably. Maybe get 64 'sinebanks' ('sinebank abstraction/module') and patch them up with various kinds of 'blocks/abstractions/modules'? What do you think? Anyone up to it? Maybe 16 (and less [finer resolution], user-defined, depending on CPU) 'sine banks' of 2048 (or more, user-defined, depending on CPU)? How hard/easy could it be? Free, if for a suggested $149 donation (--price-match-- whatever Sumu's going for)... if one is motivated by such things.

I mean, what are most if not all instruments we use hereon but rips or remixes of what came before and/or what is, maybe in large part by student/taxpayer-funded university research?

The results could of course be fed back into Sumu, and locked back in, for a price.

i already work with such a, well how should one call it, i am the only one who works with it, and it will be at some point released, it is module for a soft modular.
64 partials, with per partial FM/PM/AM, and ADSR, so you can FM a partial with another partial, or from another module. is a great thing. but still not ready, because the developer and i are lazy.

and Music Developments Syne is there, to tickle your partials....

it is already there, Reaktor Ensembles, with 1024 partials etc.

if you don't like SUMU, which is strangely nót a capital offence, nót yet i must add? or not? in this changing world. but for now.. you are free, to know thouself; which of course always means; now your place.......

[WasteLand]

I see the dumb troll is still poking about.

Has Sumu had any optimizations?

a bit, or more than a bit, there was something with the drawing of the scope and the GUI, that has been solved, it does not add to the CPU use anymore..

optimizations are served in 1.0.1, according to the email of Madronalabs, and as said several times in the Discord server.

optimizations are really interesting of course, when all is implemented, the whole code, must be 'combed'....

[Sackbut]

'Reverse Engineering' Sumu?

Could Sumu be somehow duplicated, echoed or improved in Plugdata? Probably. Maybe get 64 'sinebanks' ('sinebank abstraction/module') and patch them up with various kinds of 'blocks/abstractions/modules'? What do you think? Anyone up to it? Maybe 16 (and less [finer resolution], user-defined, depending on CPU) 'sine banks' of 2048 (or more, user-defined, depending on CPU)? How hard/easy could it be? Free, if for a suggested $149 donation (--price-match-- whatever Sumu's going for)... if one is motivated by such things.

I mean, what are most if not all instruments we use hereon but rips or remixes of what came before and/or what is, maybe in large part by student/taxpayer-funded university research?

The results could of course be fed back into Sumu, and locked back in, for a price.

i already work with such a, well how should one call it, i am the only one who works with it, and it will be at some point released, it is module for a soft modular.
64 partials, with per partial FM/PM/AM, and ADSR, so you can FM a partial with another partial, or from another module. is a great thing. but still not ready, because the developer and i are lazy.

and Music Developments Syne is there, to tickle your partials....

it is already there, Reaktor Ensembles, with 1024 partials etc.

I posted a link from the Plugdata thread to my comment about it, vis-a-vis Sumu, here.

I am aware of MD's Syne, but is it, and yours, FLOSS (Free/Libre/Open Source)?

Syne's $50 and doesn't appear FLOSS either-- neither Reaktor. The point about my mention of Plugdata is that it is FLOSS, and fits in a DAW.

Have you looked at Plugdata for your synth? Is that where you're developing it? If so, and if you're lazy as you say, would you consider outsourcing some of its development? Or...

Have you, or others reading this, leveraged AI for the coding-- maybe Windsurf, Cursor and/or

[Sackbut]

Has Sumu had any optimizations?

Hang in there. Randy will likely eventually get around to it.

[kraster]

'Reverse Engineering' Sumu?

Could Sumu be somehow duplicated, echoed or improved in Plugdata? Probably. Maybe get 64 'sinebanks' ('sinebank abstraction/module') and patch them up with various kinds of 'blocks/abstractions/modules'? What do you think? Anyone up to it? Maybe 16 (and less [finer resolution], user-defined, depending on CPU) 'sine banks' of 2048 (or more, user-defined, depending on CPU)? How hard/easy could it be? Free, if for a suggested $149 donation (--price-match-- whatever Sumu's going for)... if one is motivated by such things.

I mean, what are most if not all instruments we use hereon but rips or remixes of what came before and/or what is, maybe in large part by student/taxpayer-funded university research?

The results could of course be fed back into Sumu, and locked back in, for a price.

Good luck with that!

Sumu uses the Loris library which was developed with Kyma in mind.

It's a propietary tech (Reassigned Bandwidth-Enhanced Additive Sound Model) that simply put uses a combination of amplitude, sines and a noise element.

One of the problems with resynthesis is that inharmonic content like noise is extremely costly when using exclusively sine based FFT.

In Sumu you can see this where the partial map is divided into amplitude, pitch and noise.

At any one time you have effectively 64 "streams" of sines and noise with FM/AM.

Did Randy Jones from Madrona do something to you because you have a seriously weird obsession with trying to discredit his work?

Sumu is very transparent in how it works. It's a fantastic implementation of what is usually the most unwieldy form of synthesis.

Additive resynthesis usually yields huge amounts of data and it can be very difficult to come up with a system that allows you to interact with that data in a way that doesn't involve tedious manipulation of individual sine waves and envelopes.

And that's the thing, I don't like Sumu because it's some kind of naval gazing academic exercise. I like it because because it's a very well conceived implementation of a specific type of synthesis that is usually resistant to being user friendly.

[whyterabbyt]

Has Sumu had any optimizations?

Hang in there. Randy will likely eventually get around to it.

Im sure you'll beat him to the punch with PlugData blocks/abstractions/modules though.

[Sackbut]

The Fog Ain't Real Outside Of Our Nonexistent Gardens

'Reverse Engineering' Sumu?

Could Sumu be somehow duplicated, echoed or improved in Plugdata? Probably. Maybe get 64 'sinebanks' ('sinebank abstraction/module') and patch them up with various kinds of 'blocks/abstractions/modules'? What do you think? Anyone up to it? Maybe 16 (and less [finer resolution], user-defined, depending on CPU) 'sine banks' of 2048 (or more, user-defined, depending on CPU)? How hard/easy could it be? Free, if for a suggested $149 donation (--price-match-- whatever Sumu's going for)... if one is motivated by such things.

I mean, what are most if not all instruments we use hereon but rips or remixes of what came before and/or what is, maybe in large part by student/taxpayer-funded university research?

The results could of course be fed back into Sumu, and locked back in, for a price.

Good luck with that!

Sumu uses the Loris library which was developed with Kyma in mind.

It's a propietary tech (Reassigned Bandwidth-Enhanced Additive Sound Model) that simply put uses a combination of amplitude, sines and a noise element.

One of the problems with resynthesis is that inharmonic content like noise is extremely costly when using exclusively sine based FFT.

In Sumu you can see this where the partial map is divided into amplitude, pitch and noise.

At any one time you have effectively 64 "streams" of sines and noise with FM/AM.

Did Randy Jones from Madrona do something to you because you have a seriously weird obsession with trying to discredit his work?

Sumu is very transparent in how it works. It's a fantastic implementation of what is usually the most unwieldy form of synthesis.

Additive resynthesis usually yields huge amounts of data and it can be very difficult to come up with a system that allows you to interact with that data in a way that doesn't involve tedious manipulation of individual sine waves and envelopes.

And that's the thing, I don't like Sumu because it's some kind of naval gazing academic exercise. I like it because because it's a very well conceived implementation of a specific type of synthesis that is usually resistant to being user friendly.

The naval-gazing was/is about You Tube videos, not Sumu.

Sumu, as I already said, seems fine-- nothing really against it or Randy per se (although he did announce it way too early)-- but I've yet to hear anything that makes it 'ground-breaking'-- quite the opposite. That's what's funny.

That said, as a related aside, I'm also coming from a gift-economy perspective, so that's going to colour my interpretations.
Crony-capitalism isn't working, can't work, and is a dead man walking anyway. So tying anything into it, like synths, seems fundamentally self-defeating, even in a grab-what-you-can-while-you-can sense.

Randy likely knows this.

Lastly, AI seems to be closing the software-development gap in some regards so that apparently, we appear to be on the cusp of something like 'natural language' development. I just posted an example video in my previous comment about it. Ideally, it's supposed to accelerate the process of development and lend itself more to accessibility (such as to non-coders or casual coders), but who knows...

Perhaps Randy/Madrona Labs and Plugdata, etc., can and will leverage it... before we die or have to get back to our gardens, whichever comes first...

[whyterabbyt]

That said, as a related aside, I'm also coming from a gift-economy perspective, so that's going to colour my interpretations.
Crony-capitalism isn't working and is a dead man walking.

Hello Borbolactic.

[Sackbut]

Has Sumu had any optimizations?

Hang in there. Randy will likely eventually get around to it.

Im sure you'll beat him to the punch with PlugData blocks/abstractions/modules though.

Right. How long did Sumu take?

What's your icon a depiction of, BTW? Some sort of hare?

[Sackbut]

That said, as a related aside, I'm also coming from a gift-economy perspective, so that's going to colour my interpretations.
Crony-capitalism isn't working and is a dead man walking.

Hello Borbolactic.

Hello-o-o-o, how are you? What are you using these days?

If we have a chat with AI about (developing/developing in) Plugdata, what do you think it will know about it, if anything? Or Sumu?

[Gamma-UT]

So, what's the deal here? Create a new identity/handle to stalk different devs? Or are you just going to swap between the two?

Who's next, I wonder?

[Sackbut]

So, what's the deal here? Create a new identity/handle to stalk different devs? Or are you just going to swap between the two?

Who's next, I wonder?

Ah, the usual suspects...

I reasonably and responsibly retired 'Borbolactic' and haven't used it since (and won't use it, as that would be sockpuppetry, right?), as is my right.
I actually initially wanted Sackbut as a nickname, but at the last minute didn't think it was available-- you know, on a music site?

In retrospect, it was a good move to avoid the 'usual suspects' in my Dream Synth Redux (AKA Plugdata).

To get back to the topic, are you using Sumu and if so, how do you like it? Any samples to take a listen to?

[Gamma-UT]

To get back to the topic, are you using Sumu and if so, how do you like it? Any samples to take a listen to?

I do use Sumu. I do like it. Not getting into that game – I know how it runs and I have zero interest in trying to impress idiots.

I mean, let's face it. You don't like how this one sounds but are amazingly interested in recreating it by other means, presumably believing that the mechanics of the implementation make a huge difference to FM-able partials modulated by a 3D envelope system.

[Sackbut]

To get back to the topic, are you using Sumu and if so, how do you like it? Any samples to take a listen to?

I do use Sumu. I do like it. Not getting into that game – I know how it runs and I have zero interest in trying to impress idiots.

I mean, let's face it. You don't like how this one sounds but are amazingly interested in recreating it by other means, presumably believing that the mechanics of the implementation make a huge difference to FM-able partials modulated by a 3D envelope system.

We don't have to like what some people-- including Madrona Labs, itself-- are dishing out with Sumu, do we?
We also have the right to take the conversation in different directions with Sumu if it's about it, yes? If not, and this is a Sumu cheerleading echo chamber only, kindly let me know and I'll be on my way.

I'm interested in part in additive synthesis, and do like many of them. That's why it's curious why Sumu bucks the trend (so far at least). I was also waiting for quite awhile for it. Where's my discount? Time is money! ;P

But I have a hypothesis: It's doing wave-folding. So it's 'doing additive different than additive' and is, in a sense, not quite additive. Generate apparently does wave-folding too, as does Cherry Audio's Sines.

[Gamma-UT]

You could y'know, just read the manual and use the demo. Then you wouldn't have to guess what it does. It's not like it's some closely guarded secret.