What's the future of analog modeled plugins?
- KVRist
- 62 posts since 6 Sep, 2021
Oh, you meant ‘analog modelled plugins’...never mind
Last edited by takaperry on Thu Sep 23, 2021 7:23 am, edited 1 time in total.
multi-platinum music producer / songwriter
https://www.takaperry.com
https://www.takaperry.com
- KVRAF
- 7749 posts since 13 Jan, 2003 from Darkest Kent, UK
Yeah, I had that thought watching a certain product demo earlier this week. At this stage it would still need sanity checking, tweaking by a human etc but that will inevitably get AIed too at some point.
- KVRAF
- 15008 posts since 26 Jun, 2006 from San Francisco Bay Area
-
- KVRist
- 95 posts since 22 May, 2019
Hopefully, the future of analog emulation will be towards less bullshit. Thanks to guys like Paul Third, maybe users will become more aware of some plugin manufacturers who just put "analog emulation" labels, or famous analog brands certification, on their products while their product has nothing of the true analog behaviour. And then plugin designers will stop labeling as "analog" emulation their 100% digital linear algorithm.
Regarding machine learning, I think its usefulness could be in the algorithm design process: instead of modeling the behaviour of each component and their interactions by brute force, machine learning can help by collecting a lot of data from the real machine, and classify components and their interaction in terms of importance of their effect on the sound. Then the algorithms can be tuned to model more precisely what's important, and less precisely (or not at all) the less important components.
Incorporating machine learning in the final product itself, I don't think it would be really useful, maybe just a gimmick. Or at least, not for analog emulation.
Convolution seems like one of the best ways to truly emulate analog gear, as it's simply capturing the behaviour of the real thing, with all of its non linearities and interactions, but it's too demanding on computer resources. Maybe there's some room for progress here, with faster/lighter convolution algorithms in the future.
Regarding machine learning, I think its usefulness could be in the algorithm design process: instead of modeling the behaviour of each component and their interactions by brute force, machine learning can help by collecting a lot of data from the real machine, and classify components and their interaction in terms of importance of their effect on the sound. Then the algorithms can be tuned to model more precisely what's important, and less precisely (or not at all) the less important components.
Incorporating machine learning in the final product itself, I don't think it would be really useful, maybe just a gimmick. Or at least, not for analog emulation.
Convolution seems like one of the best ways to truly emulate analog gear, as it's simply capturing the behaviour of the real thing, with all of its non linearities and interactions, but it's too demanding on computer resources. Maybe there's some room for progress here, with faster/lighter convolution algorithms in the future.
-
- KVRAF
- 1863 posts since 11 Apr, 2008
The future is more marketing BS and this is where the most of money will be put on.
We reached that point where software is equally good as analog hardware, often even better and if somebody wants analog, it's possible to get it for a reasonable price these years (well, not everything but we don't need $50k compressor to make music these days).
So devs who still want to sell us "analog emulations" have to work hard to invent new buzz words, hype and made-up reasons to convince us that "this emulation of la2a really sounds better than la2a released 2 weeks ago by another developer and zillion other emulations". Because most people will not be able to hear the difference.
We reached that point where software is equally good as analog hardware, often even better and if somebody wants analog, it's possible to get it for a reasonable price these years (well, not everything but we don't need $50k compressor to make music these days).
So devs who still want to sell us "analog emulations" have to work hard to invent new buzz words, hype and made-up reasons to convince us that "this emulation of la2a really sounds better than la2a released 2 weeks ago by another developer and zillion other emulations". Because most people will not be able to hear the difference.
- KVRAF
- 3897 posts since 28 Jan, 2011 from MEXICO
I agree with this, marketing is the name of the game in audio software.pixel85 wrote: ↑Thu Sep 23, 2021 3:36 pm The future is more marketing BS and this is where the most of money will be put on.
We reached that point where software is equally good as analog hardware, often even better and if somebody wants analog, it's possible to get it for a reasonable price these years (well, not everything but we don't need $50k compressor to make music these days).
So devs who still want to sell us "analog emulations" have to work hard to invent new buzz words, hype and made-up reasons to convince us that "this emulation of la2a really sounds better than la2a released 2 weeks ago by another developer and zillion other emulations". Because most people will not be able to hear the difference.
dedication to flying
-
- KVRAF
- 35449 posts since 11 Apr, 2010 from Germany
Analog modelled or not, I personally reached a certain "threshold", where, when I demo new stuff, I'm mostly either disappointed, or keep thinking "Well... I already got that, in equal quality". So, yeah, it's mostly marketing to keep the ball rolling. Same with DAW's. A new major version every 1 or 2 years, for what really? To have some small feature additions noone needed in the past to make music?
You have to understand the industry. And the necessity to make money. And then decide for yourself whether or not you really need new software, or if it's just for the "Oh, I got hyped!" factor. When I study the buying behavior, I see a lot of the latter (guilty of it as well, more than once...).
You have to understand the industry. And the necessity to make money. And then decide for yourself whether or not you really need new software, or if it's just for the "Oh, I got hyped!" factor. When I study the buying behavior, I see a lot of the latter (guilty of it as well, more than once...).
- KVRAF
- 23502 posts since 12 Jul, 2003 from West Caprazumia
With DAWs there is still a far way to go... but in regards to most plugins stuff I agree, yes. We're covered - tenfold.
"Preamps have literally one job: when you turn up the gain, it gets louder." Jamcat, talking about presmp-emulation plugins.
- KVRAF
- 5512 posts since 2 Sep, 2019
I think it's going to be continuous movement towards absolute accuracy, until the experience and results of using a plugin model is indistinguishable from the original hardware. Because really, that is what the market (us) wants.
Developers who do not keep up, and regularly update their existing models with higher levels of accuracy as modeling techniques improve and processing power increases, will see their market shares dwindle and their products and brand becomes obsolete.
We will also see an ever-expanding selection of modeled gear, with increasingly eclectic, novel, and obscure gear being modeled.
All of this will keep developers busy for the foreseeable future.
Developers who do not keep up, and regularly update their existing models with higher levels of accuracy as modeling techniques improve and processing power increases, will see their market shares dwindle and their products and brand becomes obsolete.
We will also see an ever-expanding selection of modeled gear, with increasingly eclectic, novel, and obscure gear being modeled.
All of this will keep developers busy for the foreseeable future.
THIS MUSIC HAS BEEN MIXED TO BE PLAYED LOUD SO TURN IT UP