Plug-In Development Frontiers

[Unaspected]

The ultimate synth/effect in 1000 years from now would be a plugin that has one button. You push that bottom and it creates an new track or sound for you that matches your thoughts, that is absolutely aaaawchesome, that has never been heard before and that will be on the top charts. every time.

But then, every one can do it. so tracks wont sell. and so the plugin wont sell. ends up being free. And then some one will come and say. "hey, this is boring, I wanna be involved.". And we're back to the moog modular again.

It seems a number of people would love something like that from what I have seen in forum posts over the years - but none of those people were musicians. It could be useful for uncreative types working in marketing but no use for musicians or sound engineers. The idea of hitting a single button to create music does indeed sound quite boring and joyless.

It was meant as a joke. Apparently a bad one

Hehe. If you'd seen the things that I've seen.

We're obviously in the realms of fantasy here but it could be useful as a writing tool. Though I've always found that the act of picking up a guitar or reaching for my controller keyboard leads to developmental ideas that I might never have thought up without the physical interaction.

[S0lo]

I've always found that the act of picking up a guitar or reaching for my controller keyboard leads to developmental ideas that I might never have thought up without the physical interaction.

Yup, same here

[JCJR]

Ok I see what you're saying. That's not really related to what people like though which is what I was talking about. Soundmodel was wondering why dance musicians make similar style music over and over. To say it's down to a lack of imagination, creativity or lack of new technological advances is naiive. Song writers will write what people like and what people like is often simple and familiar.

"Big Markets" of people seem to like the same thing, same genre. Imagine a population bell curve of people who respond to "soul-type R&B" and superimposed population bell curves of people who respond to electronic or acoustic irish or whatever.

Thing is, a composer who doesn't "like" or "understand" a genre probably wastes his time trying to compose for that genre. Seems a musician must "believe" what he is selling or the audience can too often, too easily identify the work as insincere.

Music which is heavily machine-dependent and non-realtime, maybe it would be harder for the audience to know if the musician is "faking it" trying to exploit a fad. But I suspect somehow listeners could identify the cynical commercial parodies even in sequencer music.

Am just saying, you would have to be a hell of a good actor to sell a rock'n'roll set if you don't like rock'n'roll. And you would have to be a hell of a good actor to sell a techno-dance song if you don't like techo-dance, or sell a surf song if you don't like surf music.

Even the successful genre stars tend to be really good actors IN ADDITION to really liking the genre they are selling. So to be successful in a genre you don't like, would require near super-human acting ability.

Am just saying, maybe a lot of composers/musicians play disco if disco is the only thing selling, or whatever. But only the musicians who REALLY LIKE DISCO will reliably connect to the listeners. You can't sell an Elvis song to an audience unless you genuinely like Elvis. Etc.

[low_low]

One of the frontiers is for VST manufacturers to start using computer graphics cards for their DSP parallel processing. We've already hit CPU limits on some VST's such as Diva, forcing users to limit the number of unison voices, etc. It's often misunderstood because it isn't an ACTUAL cpu limit (look at your cpu graph and compare that to your DAW cpu graph and know the truth), it's a limit on how responsive the computer is when operating in near real time with low latency, but graphics card processing would still probably help.

[Ananke]

"Big Markets" of people seem to like the same thing, same genre. Imagine a population bell curve of people who respond to "soul-type R&B" and superimposed population bell curves of people who respond to electronic or acoustic irish or whatever.

Thing is, a composer who doesn't "like" or "understand" a genre probably wastes his time trying to compose for that genre. Seems a musician must "believe" what he is selling or the audience can too often, too easily identify the work as insincere.

Music which is heavily machine-dependent and non-realtime, maybe it would be harder for the audience to know if the musician is "faking it" trying to exploit a fad. But I suspect somehow listeners could identify the cynical commercial parodies even in sequencer music.

Am just saying, you would have to be a hell of a good actor to sell a rock'n'roll set if you don't like rock'n'roll. And you would have to be a hell of a good actor to sell a techno-dance song if you don't like techo-dance, or sell a surf song if you don't like surf music.

Even the successful genre stars tend to be really good actors IN ADDITION to really liking the genre they are selling. So to be successful in a genre you don't like, would require near super-human acting ability.

Am just saying, maybe a lot of composers/musicians play disco if disco is the only thing selling, or whatever. But only the musicians who REALLY LIKE DISCO will reliably connect to the listeners. You can't sell an Elvis song to an audience unless you genuinely like Elvis. Etc.

I'd agree with you but I wasn't suggesting that the people making music were faking it or not really interested in whatever genre they were creating for (although I'm sure that exists to some extent).

My original point was that I don't think there is a massive driving need from the music industry for more sophisticated software to create more ambitious music. Most popular music is created easily with the tools that are currently around. I think that any kind of new "frontiers" reached are mainly from the academic community.

[Richard_Synapse]

One of the frontiers is for VST manufacturers to start using computer graphics cards for their DSP parallel processing. We've already hit CPU limits on some VST's such as Diva, forcing users to limit the number of unison voices, etc. It's often misunderstood because it isn't an ACTUAL cpu limit (look at your cpu graph and compare that to your DAW cpu graph and know the truth), it's a limit on how responsive the computer is when operating in near real time with low latency, but graphics card processing would still probably help.

GPUs are fantastic if you work on stuff ideally suited to their specific structure. But DSP algorithms and the required (very) low latency are problematic for GPUs. For the near future the Intel i9 is ihmo the way to go.

Richard

[S0lo]

One of the frontiers is for VST manufacturers to start using computer graphics cards for their DSP parallel processing. We've already hit CPU limits on some VST's such as Diva, forcing users to limit the number of unison voices, etc. It's often misunderstood because it isn't an ACTUAL cpu limit (look at your cpu graph and compare that to your DAW cpu graph and know the truth), it's a limit on how responsive the computer is when operating in near real time with low latency, but graphics card processing would still probably help.

GPUs are fantastic if you work on stuff ideally suited to their specific structure. But DSP algorithms and the required (very) low latency are problematic for GPUs. For the near future the Intel i9 is ihmo the way to go.

Richard

I think one of the main and probably stubborn problems to crack is that most audio processing (with a few exceptions like polyphony) is not parallelizable. Which is where allot of the new advances in hardware processing power is going. Serial processing has to improve.

[vortico]

Oh, one more thing to add. We're in the DSP board, but there's a lot of innovated needed in the hardware controller market. Still more things to try like haptic interfaces, modular knobs/buttons/cable MIDI controllers, expressive keyboards with MPE, and electro-physical sound making devices.

[JCJR]

I'd agree with you but I wasn't suggesting that the people making music were faking it or not really interested in whatever genre they were creating for (although I'm sure that exists to some extent).

My original point was that I don't think there is a massive driving need from the music industry for more sophisticated software to create more ambitious music. Most popular music is created easily with the tools that are currently around. I think that any kind of new "frontiers" reached are mainly from the academic community.

Hi Ananke. I agree. Maybe some developers can imagine some unique need lots of generic customers would want once the new thing is created, though the developer does not see a personal need for the new tool. Identify a new niche and then fill it out of profit motive.

I was never imaginative that way. Sometimes I wrote tools I didn't necessarily need but had seen people ask for, which seemed a good idea. Otherwise, just developed things I personally needed and was either lucky or not depending on whether enough other people thought it a good enough idea to buy into. I think new tools that are easy to explain/demo, or self-evidently "useful" have the better odds. If you have to write long white papers or create hours of youtube tutorials just to explain what the tool is good for, then it is probably an uphill slog trying to sell it.

Maybe explains the plethora of chorus units, compressors and EQs. Lots of people already know how to use such tools. Not a big educational effort to convince the user that he needs a compressor or EQ. "See, I got this shiny new shovel fer sail. It works just like any other shovel but better. Has a nice paint job, plenty of chrome and a catchy brand name. This here long list of testimonials from famous ditch diggers who think my shovel is the best."

Apologies was not trying to be preachy about "ya need to like the music before you can sell it." I just think it is approximately accurate and it took awhile to learn the lesson, so was trying to pass it on.

The main consequence is that even really good musicians can be "commercially doomed by their elevated tastes" if the tastes are more effete than those of potential audiences.

Maybe some year, worldwide sales of house, rap, hard bop, metal, country have all fallen so low as to be lost in rounding error, The only thing selling is the new Swine Swill Beat sweeping the charts in all known galaxies. So some real talented musician who hates Swine Swill music says to himself, "Self, If you are gonna sell any records this year then you better record Swine Swill music regardless how bad it sucks". But somehow the audience can tell, and they don't buy the records anyway.

Then there is some young musician who loves Swine Swill music even better than the fans. That is the kind of fella who will sell records.

On the other hand, after the Swine Swill Beat fashion has faded, unless the young Swine Swill star has tastes wide enough to appreciate newer fashions, then he can only hope that his fans live long enough and remember good enough so that he can keep playing Swine Swill Revival oldies tours well into his dotage.

[Benutzername]

I think one of the frontiers will be artificial intelligence. Maybe not in the DSP part of the software but in the creative, editing and mixing stages.

There are already song generators out there but they just follow fixed schemes. A real AI will go much further. Just feed it with your Spotify stream and your previous work and it will learn what you like and learn what you don't like.

This AI would allow to turn a simple loop into a complete song with a just a single mouse click. Just enter the genre or a reference track and the AI will analyze the current state, built a matching song structure and automatically fill in the gaps.

Editing would be a breeze as the audio editor knows that it is working on a snare and it knows how it has to sound to fit in the target mix. No more manual cutting and no need to fiddle with thresholds, tunings and eq bands anymore.

Same with mixing. If you stay in a specific genre then mixing is usually following certain rules with just minor tweaks. Also the type of plugins are more or less the same. The AI could detect the genre, the content of the tracks, the overall density and the musical progressions and build a rough mix that just needs a bit of fine tuning.

And the AI will be getting better every time because it learns from the manual changes that you make to it's suggestions. Every new song will need less tweaking.

[JCJR]

Such a level of AI might be interesting. Timescale is problematic. I don't know how to estimate. Both high-level AI and practical fusion power have been "a decade away" since the 1950's. They might still be "real soon now" in 2050, dunno. Or maybe next week?

There are priorities. Given a choice between intensive development of AI music, rocket cars or jet packs, I wish they would focus on Louis Wu's Autodoc (Niven's Ringworld series and others). Deliver a decent nanotech autodoc before its too late and I'll somehow manage to get by without AI or jet packs.

[Urs]

In my opinion, academia does not often push new boundaries in our field.

As long as we live in a nostalgia-driven market, it's old farts with experience who figure it out, not young and ambitious lads who seek the new by all means. And that won't change because of the duality/dilemma of software vs. hardware. Unlike similar fields like Photography where digital was "merged into" the field with benefits, audio hardware is stubbornly setting the bar (and that's not remotely justified by any result).

[bbtr]

In my opinion, academia does not often push new boundaries in our field.

As long as we live in a nostalgia-driven market, it's old farts with experience who figure it out, not young and ambitious lads who seek the new by all means. And that won't change because of the duality/dilemma of software vs. hardware. Unlike similar fields like Photography where digital was "merged into" the field with benefits, audio hardware is stubbornly setting the bar (and that's not remotely justified by any result).

That's, like, an overcomplification.

[vortico]

Both high-level AI and practical fusion power have been "a decade away" since the 1950's. They might still be "real soon now" in 2050, dunno. Or maybe next week?

True for fusion (I came from that field), but machine learning for modulation-rate music has been ready for at least 5 years. The problem is that it's near impossible to make something that 1) is musical at all, 2) gives users the proper balance of control while making its own helpful decisions, and 3) appeals to the market (which is mostly users with super conservative workflows) enough to fund research costs. Will you be able to go to a club with a musician performing a neural network? Probably not, subtractive synths from 1970 are just fine.

This isn't primarily the fault of the developers' skills or the customers' opinions, it's just that machine learning is not a correctly-shaped tool for this field. I have faith that someone can make it work to excite us experimenters, but it's still a solution looking for a problem. For the general public it will be like using a helicopter to drop your kids off at school. Needlessly unnecessary, over-extravagant, and doesn't really improve quality of life.

Agree with Urs that academia isn't a huge factor in pro audio product offerings. Companies would be unaffected if academic audio research halted, so you have to realize that stimulating commerialism is not the purpose of academic research. They are doing it for the benefit of increasing the body of human knowledge, not to allow new products to enter the market in 3 years. That will happen regardless of new research.

[Aleksey Vaneev]

The problem AI won't work beside being just a more complex "curve fitting" technology is that human reality is driven by hype and money. And that's the realm of "magic" or unconscious, not really ever studied well. The closes scientific approach is The Global Consciousness Project http://noosphere.princeton.edu/

If you meditated on it a bit, you'll understand that consciousness creates NEW reality while AI can only reproduce/remix OLD reality. That's why AI will never give any real competitive advantage. AI or solving system of equations is great for optimizing OLD reality, but when balances in human psyche shift, they quickly become outdated.

Simply put, AI can't be a celebrity and that makes its creative potential very small. Well, it can be probably used to produce "quirks", but these "quirks" must be then refined by a human with creative potential.

If you ever delved into "AI mythology", you may have got an idea that humans as a collective have very strong "natural" protection against AI technology. That's what puts us aside from animals - animals can be controlled by AI while humans are controlled by their own collective. So, be it domestic AI or AI from cosmos, it can't easily control humans. AI's "signals" will tend to fail the collective reality check of humans.