RayBlaster - a radically different approach to sound generation.

[robelanator]

To avoid inadvertently starting a flamewar with my very first post here, I suppose I should provide my definition of what “musically useful” means since it can be a subjective term.

For my purposes, it would mean tones that can be played as a melody or combined into chords that would be recognizable as such to the average layman listener of music within the traditional Western genres.

Thanks again!

[Markus Krause]

What is the max polyphony and memory usage per instance?

The first instance needs around 200 MB. Further instances need less as they can share resources in the memory.

Polyphony can go up up to 256 (64 voices playing with 4x unsion).

Does Rayblaster's randomizer give "musically useful" results?

Yes. Unlike most other synthesizers it does not just set random values for each parameter. It uses a very smart algorithm with a statistic approach like Warlock. Today's hype-marketing would sell it as 'Artificial Intelligence'.

There is a free demo which you can download here for testing:
https://www.tone2.com/download.html

To create a random patch click 'Edit'->"create random patch"

[robelanator]

Thank you, Markus. I will download the demo and give it a try later this week.

[rezoneight]

Yes. Unlike most other synthesizers it does not just set random values for each parameter. It uses a very smart algorithm with a statistic approach like Warlock. Today's hype-marketing would sell it as 'Artificial Intelligence'.

Now now Markus, you know its Machine Learning and not AI...

[chk071]

Which is a similar hype-marketing buzz word.

Even when it's "learning", a machine can't act in any different way than it's supposed, pre-defined and programmed way. Anything else is sci-fi fairy tale stuff really.

[BBFG#]

Which is a similar hype-marketing buzz word.

Even when it's "learning", a machine can't act in any different way than it's supposed, pre-defined and programmed way. Anything else is sci-fi fairy tale stuff really.

More people are finally realizing that "Artificial Intelligence" has nothing to really do with the product and everything to do with the niche market they're selling it to.

[martinjuenke]

Intelligence, Learning, Buzz Words: who cares?
How does it sound? That's the Bratwurst on the plate!

[rezoneight]

Which is a similar hype-marketing buzz word.

Even when it's "learning", a machine can't act in any different way than it's supposed, pre-defined and programmed way. Anything else is sci-fi fairy tale stuff really.

There is nothing predefined about ML, thats kinda the point. It's used in areas where there is no good pre-defined algorithm and why you're feeding the model large datasets so it can refine things and "learn" on its own. Thats why facial recognition is so damned good these days for instance. Any buzzwordiness is...whatever. That doesn't diminish what's been accomplished via ML techniques in software.

[rezoneight]

Intelligence, Learning, Buzz Words: who cares?
How does it sound? That's the Bratwurst on the plate!

Companies like Izotope and Baby Audio care Audio is most certainly an area where ML/AI techniques can be used very effectively. Creating presets probably isn't one of them. That seems algorithmically relatively easy to accomplish.

[vurt]

Which is a similar hype-marketing buzz word.

Even when it's "learning", a machine can't act in any different way than it's supposed, pre-defined and programmed way. Anything else is sci-fi fairy tale stuff really.

one of them won't rise up and destroy mankind.
sadly, we don't know which

[chk071]

Which is a similar hype-marketing buzz word.

Even when it's "learning", a machine can't act in any different way than it's supposed, pre-defined and programmed way. Anything else is sci-fi fairy tale stuff really.

There is nothing predefined about ML, thats kinda the point. It's used in areas where there is no good pre-defined algorithm and why you're feeding the model large datasets so it can refine things and "learn" on its own. Thats why facial recognition is so damned good these days for instance. Any buzzwordiness is...whatever. That doesn't diminish what's been accomplished via ML techniques in software.

It's still a programmed pattern. What else? All that "learning" is nothing but a programmed behavior.

[AnalogGuy1]

When I write a "regular" program, I define the behavior intentionally. I can program the output of x and y inputs to be x+y. I can look at the code and verify that for any x and y that it will return x+y without feeding it any examples.

If I write a ML program, I "program" it by feeding it large datasets and let it program itself. Even if most of the dataset shows are examples of x and y inputs as returning x+y, I can't be sure that this is what the ML program will do by looking at the weighting matrices that form the resulting ML program. Further, for inputs outside of the dataset, the output may be x+y generally, but on an off day might be a grapefruit.

ML is very different than typical programmed behavior.

[briefcasemanx]

Innovative new synthesis types are cool, but have you found a backdoor into mathematics itself, proving the existence of God, like at least 2 other plugin creators have?

[chk071]

If I write a ML program, I "program" it by feeding it large datasets and let it program itself.

The point is that you program it to do so. If machine learning wouldn't do what it's supposed to do then it would be completely useless.

Anyway, I really don't want to bloat this thread with offtopic.

[martinjuenke]

If I write a ML program, I "program" it by feeding it large datasets and let it program itself.

The point is that you program it to do so. If machine learning wouldn't do what it's supposed to do then it would be completely useless.

Artificial intelligence is far more advanced nowadays. There was an interesting article recently in the Neue Zuricher Zeitung for example. Crazy stuff.
But that level is not included in music plugins obviously. Therefore we should here focus on one thing: does an instrument provide us with cool and interesting sounds (no matter how they're produced)?