When do you think Logic Pro 11 will come out?

[andypryce]

I'm not sure how a neural network can help, unless you want it to learn your style of music and compose it for you. I would find that creatively stifling. I prefer using aleatoric methods when stuck for an idea.

Would the neural network be useful for resynthesis of audio (Alchemy), timestretching improvements, Flex Pitch improvements?

Maybe we will get Logic's own Vocaloid

[syntonica]

I'm not sure how a neural network can help, unless you want it to learn your style of music and compose it for you. I would find that creatively stifling. I prefer using aleatoric methods when stuck for an idea.

Would the neural network be useful for resynthesis of audio (Alchemy), timestretching improvements, Flex Pitch improvements?

It could, but what would it learn against? Maybe it ciuld do live 5 versions and let the human pick the best? We're also getting into "crowdsourcing" land here and data could be pooled from multiple users, like u-he's Protoverb. For better or worse.

If you look at AI art made by programs that process tons of art, they are still really weird looking. It's an interesting question though, as it's dealing with a specific granular (no pun intended) task rather that composing a full piece of music. But I, personally, am more comfortable with AI helping to shape tools and not putting more humans out of work.

Sorry for the disjointed reply... it's been a very long day of me swearing at my programming computer.

[medienhexer]

Huh? You said that your m1 was hardly getting you 10 tracks. That is something I can hardly imagine. Unless you run into RAM issues due to Rosetta.

Or maybe you said something else but phrased it VERY confusingly. A 2014 Macbook was never part of the discussion.

[pdxindy]

I'm not sure how a neural network can help, unless you want it to learn your style of music and compose it for you. I would find that creatively stifling. I prefer using aleatoric methods when stuck for an idea.

Would the neural network be useful for resynthesis of audio (Alchemy), timestretching improvements, Flex Pitch improvements?

It could, but what would it learn against? Maybe it ciuld do live 5 versions and let the human pick the best? We're also getting into "crowdsourcing" land here and data could be pooled from multiple users, like u-he's Protoverb. For better or worse.

If you look at AI art made by programs that process tons of art, they are still really weird looking. It's an interesting question though, as it's dealing with a specific granular (no pun intended) task rather that composing a full piece of music. But I, personally, am more comfortable with AI helping to shape tools and not putting more humans out of work.

Sorry for the disjointed reply... it's been a very long day of me swearing at my programming computer.

I was wondering if it could try a whole bunch of say resynthesis variations and check which is closest to the original... or something like that. I know nothing how that works, but I have the idea of some iterative process where it get as close as possible to an objective. Maybe that is all wrong.

[medienhexer]

And I do think that ML could improve time stretching and pitch correction. A model could be trained to recognize the instrument playing and follow each instrument‘s pitch in a mixed signal.

You could train typical unwanted noise like wind, streets, sirens, etc. and remove them. Actually, ML is probably the only way forward right now.

[syntonica]

I was wondering if it could try a whole bunch of say resynthesis variations and check which is closest to the original... or something like that. I know nothing how that works, but I have the idea of some iterative process where it get as close as possible to an objective. Maybe that is all wrong.

That's more or less how it works. It just needs something as a target to model. When it comes to neural networks producing "original" works of art, then I'm lost in the weeds as to how that works.

[pdxindy]

I was wondering if it could try a whole bunch of say resynthesis variations and check which is closest to the original... or something like that. I know nothing how that works, but I have the idea of some iterative process where it get as close as possible to an objective. Maybe that is all wrong.

That's more or less how it works. It just needs something as a target to model. When it comes to neural networks producing "original" works of art, then I'm lost in the weeds as to how that works.

Even original works of art are still all derivative. I don't have much interest in that area cause art, or any creative task is more about the process and ones learning as a result. Programming computers to accurately mimic human creations loses the point of its value and becomes just a way to produce more commercial products. It might lead to humans seeing what is the real value because the value of the physical/visual result will be diminished.

Logic's Flex Pitch is not as good as Melodyne. I mean the resulting quality is not quite as good. So I thought maybe that would be a good task to put it to.

[Johnnyjohn]

I still believe the neural engine could be pretty useful in recognizing repeating patterns by the user.
For example me as an electronic artist myself do a lot of sample copy pasting for build ups, would be incredible useful to get a suggestion feature to switch to 8th/16th/32th on various rhythms as soon as the engine recognizes such behavior.
Another cool feature would be to learn and suggest various chord patterns or plugins that usually follow after action x.

[lightsfadelow]

One additional thing I’d like to see is an updated 3rd party AU instrument spec that would allow 3rd parties full preset libraries to be visible in the Logic library view. (And let’s make the library search/filter/index much more full features.) imagine not having to hunt through multiple plugins’ different preset browsers to find that perfect sound you know has to exist somewhere, but you’re not sure which of the 50 instruments it’s in.

[jules99]

One additional thing I’d like to see is an updated 3rd party AU instrument spec that would allow 3rd parties full preset libraries to be visible in the Logic library view. (And let’s make the library search/filter/index much more full features.) imagine not having to hunt through multiple plugins’ different preset browsers to find that perfect sound you know has to exist somewhere, but you’re not sure which of the 50 instruments it’s in.

That would be amazing. Or better yet, have some kind of way to access to NKS presets. Imagine being able to preview any preset from that many more virtual instruments.

[Johnnyjohn]

10.7.7 is out folks - still nothing major

https://support.apple.com/en-lamr/HT203718

[lightsfadelow]

Oh thank god they fixed the Flex Pitch issue1

[chagzuki]

I'm still on 10.7.4 because these minor updates aren't solving the issue I have with projects taking ludicrously long to load (since 10.7.5).

[Johnnyjohn]

bump

[Johnnyjohn]

should arrive shortly