Any one try the AnalogX AI stuff yet?

[Ed DeGenaro]

Especially their Genesis plugin?
Seems promising, about to make coffee and will check it out.

AnalogX

 

[2dor]

I do recall seeing some mentions in the NAM Facebook group on profiling studio gear.

I do wonder what they did about handling the compression stuff; out of the box, NAM isn't built to profile compression.

 

[Ed DeGenaro]

Whatever they did I dug there demo enough that I actually just paid for the full version

 

[ian_dissonance]

I remember when I asked about using NAM for exactly this purpose with TZ3K and someone who pretended to be smarter me said it was a bad idea. There’s no reason NAM can’t profile studio hardware just like it does guitar amps and this seems to be the proof. Of course NAM can profile compression. Does compression not happen inside an amp circuit? Isn’t profiling essentially putting a known quantity through a signal chain and extracting the difference to apply elsewhere?

 

[2dor]

I remember when I asked about using NAM for exactly this purpose with TZ3K and someone who pretended to be smarter me said it was a bad idea. There’s no reason NAM can’t profile studio hardware just like it does guitar amps and this seems to be the proof. Of course NAM can profile compression. Does compression not happen inside an amp circuit? Isn’t profiling essentially putting a known quantity through a signal chain and extracting the difference to apply elsewhere?

I've read all sorts of stuff about what it can and can't do; haven't tried it for anything else aside from amps and boost pedals.
But I guess for what some pres would do, it should work fine. Not sure about more complex chains and time-based stuff; Steve did a P.o.C. where he NAM-ed a Tremolo circuit last April so it could be done just not out of the box I guess.

 

[[Nathan]]

I had a play around with it. For now its just another piece of tech I'll watch and wait to see, but don't have the time to be part of being on the front line of messing around with it. My simplistic brain thinks of NAM files as trained neural networks to match the source (amps/analog gear) at a fixed set of settings. To me that's perfect for amp captures or preamp captures or anything in the realm of set and forget. When it comes to full amp controls and switches as well as studio equipment with adjustable settings like a compressor it conceptually makes less and less sense to me.

I know it's possible to capture compression in the sense of a compressor, I know its possible to stitch multiple NAM captures together but at what point is modelling algorithms more beneficial than brute forcing NAM captures? Maybe a hybrid approach where NAM is capturing one element of hardware and it's paired with time based algorithms for more precise control... I don't have the answers but they're all good questions. Happy to watch and see the tech unfold but after using AnalogX for 10-30 minutes I was like, yeah this isn't for me, I'll stick to commercial plugins for this stuff for now.

 

[MirrorProfiles]

I had a play around with it. For now its just another piece of tech I'll watch and wait to see, but don't have the time to be part of being on the front line of messing around with it. My simplistic brain thinks of NAM files as trained neural networks to match the source (amps/analog gear) at a fixed set of settings. To me that's perfect for amp captures or preamp captures or anything in the realm of set and forget. When it comes to full amp controls and switches as well as studio equipment with adjustable settings like a compressor it conceptually makes less and less sense to me.

I know it's possible to capture compression in the sense of a compressor, I know it’s possible to stitch multiple NAM captures together but at what point is modelling algorithms more beneficial than brute forcing NAM captures? Maybe a hybrid approach where NAM is capturing one element of hardware and it's paired with time based algorithms for more precise control... I don't have the answers but they're all good questions. Happy to watch and see the tech unfold but after using AnalogX for 10-30 minutes I was like, yeah this isn't for me, I'll stick to commercial plugins for this stuff for now.

Basically all of this.

I don’t really see what it can achieve that can’t be done with existing methods (which are a lot nicer and easier to control and have less drawbacks). Studio gear can be very interactive depending on how things are set and capturing those behaviours with machine learning is not going to be straightforward.

Even though it’s a different approach - look at Acustica (who sample their equipment with their own process). It’s been around for nearly 20 years and is still a nightmare to use.

I think machine learning for sure has a part to play in analog emulation, but not by brute forcing entire pieces of equipment regardless of what they do, using a guitar amp modeller. I’d love to know an instance where it seems practical or useful, at least in its current form. Just screams one step sideways, 4 steps back to me atm.

 

[ian_dissonance]

I’d guess 90% if the people interested in this and things like STL ControlHub are mostly after the “color” from a driven signal chain and not so much trying to get exactly 3.4 attack and 2.6 release on a Distressor to fit a vocal. They just want some transformer/tube mojo, or like classic set and forget buss settings.

 

[MirrorProfiles]

I’d guess 90% if the people interested in this and things like STL ControlHub are mostly after the “color” from a driven signal chain and not so much trying to get exactly 3.4 attack and 2.6 release on a Distressor to fit a vocal. They just want some transformer/tube mojo, or like classic set and forget buss settings.

There’s already plugins that can do that really accurately though - are they inadequate or lacking in some way? Are they the bottleneck on making something sound as good as it needs to? These plugin companies have decades of experience modelling gear in a variety of ways and have already done the hard work on this stuff.

How many records these days are made almost entirely ITB? Sometimes worth remembering how clean most analog gear is designed to be, and is usually used.

Feels like a lot of people are looking for cheat codes rather than listening to and responding to what the music is calling for. And if there is ever going to be some kind of technical bottleneck on someone’s work, the best thing they can improve is their acoustics and monitoring. There isn’t a tape or neve emulation on the planet that will benefit a song more than actually hearing what’s going on accurately. There’s no such thing as “set and forget” bus settings, unless every single song is written, performed and recorded exactly the same way each time.

 

[EOengineer]

I feel like many of these AI <insert-literally-anything> products are more or less a biproduct of the poor state of tech investment right now than anything else. It’s VERY difficult to attract investment unless you’ve slapped AI all over your product.

I hope we continue to see the tech progress forward, and hopefully not at the expense of white glove modeling. Luckily we have advocates in companies like Fractal, L6, and others.

 

[ian_dissonance]

There’s already plugins that can do that really accurately though - are they inadequate or lacking in some way? Are they the bottleneck on making something sound as good as it needs to? These plugin companies have decades of experience modelling gear in a variety of ways and have already done the hard work on this stuff.

How many records these days are made almost entirely ITB? Sometimes worth remembering how clean most analog gear is designed to be, and is usually used.

Feels like a lot of people are looking for cheat codes rather than listening to and responding to what the music is calling for. And if there is ever going to be some kind of technical bottleneck on someone’s work, the best thing they can improve is their acoustics and monitoring. There isn’t a tape or neve emulation on the planet that will benefit a song more than actually hearing what’s going on accurately. There’s no such thing as “set and forget” bus settings, unless every single song is written, performed and recorded exactly the same way each time.

Oh for sure, it’s not for me. I probably really only need like less than 10 plugins to get things done and that’s including a couple things for space and special FX. But there are definitely some high-level guys out there that barely touch their buss comps and EQs, they’re just leaning into them until it sounds right or the meter does the thing. Just like guitar tones, I very much prefer modeling to profiling/capturing in plugins. But, there is a whole industry built around people who want to make music with as little effort as possible and they will spend $800 a year on random plugins looking for a magic button.

 

[MirrorProfiles]

I feel like many of these AI <insert-literally-anything> products are more or less a biproduct of the poor state of tech investment right now than anything else. It’s VERY difficult to attract investment unless you’ve slapped AI all over your product.

I hope we continue to see the tech progress forward, and hopefully not at the expense of white glove modeling. Luckily we have advocates in companies like Fractal, L6, and others.

I think this is definitely something wider going on that is getting sadly noticeable. I think the influence of AI is working in both directions - lazy products from those hamfisting it into anything and everything in the hope that something is useful, and caution from others who are waiting to see what effect it has on everyone.

Part of me wondered if it was just me becoming more and more numb to new releases but it really does feel like the rate of advancement across music software has slowed RIGHT down. I think its largely down to two factors:

- with the recording industry expanding wider and wider, there is a necessity/expectation for plugins to cost $29. Some might be closer to $50 on release, some might stretch to $100 or even $200 depending on marketing and perceived value. Often it's significantly less than $29 when you factor in bundle pricing. Even though they might be selling more in volume, it's a difficult market to compete in while still offering support, updates, R&D, development costs, new features etc. Plugins can sell quite well without needing too much development cost or bringing anything new to the table.

- this has led to big established companies competing side by side with the one man developer's. While there are some awesome products designed by individuals, there's absolutely something that teams of people can accomplish that one person on their own simply can't do. There is an abundance of companies now with minimal staff, minimum DSP knowledge, minimal support/experience, churning out run of the mill plugins using basic algorithms. Bigger companies struggle to recoup on more advanced software that requires more R&D so they are also making more basic plugins that yield a quicker return and less risk.

- one other big factor is the impending influence of AI. If you have money to invest in R&D on new products, you're probably going to wait and see what's around the corner before blowing your load on something that may well be within the entire planet's grasp instantly and for next to nothing in 5 or 10 years. It's quite hard to plan too far ahead and I think generally a lot of companies are treading carefully right now.

- also add in the after effects of that weird covid boom that happened, where a lot of companies grew rapidly, sold out to private equity and now the reality sets in that many of these companies don't really have a long term strategy or big position in the market.

There's part of me that thinks the blowout pricing from companies like UAD and IK are in part due to knowing that in a few years the entire market for that stuff will have shifted on its head and this is the last period of the old era.

 

[MirrorProfiles]

Oh for sure, it’s not for me. I probably really only need like less than 10 plugins to get things done and that’s including a couple things for space and special FX. But there are definitely some high-level guys out there that barely touch their buss comps and EQs, they’re just leaning into them until it sounds right or the meter does the thing. Just like guitar tones, I very much prefer modeling to profiling/capturing in plugins. But, there is a whole industry built around people who want to make music with as little effort as possible and they will spend $800 a year on random plugins looking for a magic button.

I agree with this - I think there's kind of a separation between what goes on with the top level guys who have a Pultec or GML or whatever that they leave alone and run everything through, and the kid in their bedroom thinking that their plugin emulation of it ISN'T QUITE good enough and is making their work sound inferior to the guy they look up to.

The reality is those big names would produce exactly the same quality of work with or without any particular gear. If their Pultec is adding 3dB of top end, and the tracks that come in are already too bright, they're just going to compensate by EQing the tracks differently. Most of the time they're barely even thinking about gear because it largely gets in the way of actually doing the work. It's fun to own and can be a cool inspiring thing to have in the studio. A random ass NAM file doesn't really scratch that itch and just leads whoever's using them to have something else to doubt in their mixing chain.

But yeah, there are for sure a lot of people who are willing to throw money at all kinds of stuff in the hope that its the silver bullet and the solution to what's been holding them back. There are still COUNTLESS new plugins released of the same old gear - Pultecs, 1176's, LA2A's, Fairchilds, dbx160 etc, all as if whatever is out there isn't QUITE good enough and that this new one has some kind of minute difference that will suddenly translate into better results. It's wishful thinking but people are insecure and will think that the differences matter.

Meanwhile you have the top level guys out there mixing 3+ songs a day with old ass waves plugins not worrying about aliasing or oversampling or whether their SSL eq is more accurate than another one. They're just doing the work, getting it approved and moving on to the next one as quickly and efficiently as possible.

I have WAY too many plugins, but the reason for using different ones is for inspiration and fun. I could mix just fine with stock plugins but using 3rd party stuff keeps it interesting and inspiring. Going through a list of presets (be it in something like ControlHub) or NAM files doesn't really allow you to get into the song, its just swinging in the dark and won't really get very far.

 

[ian_dissonance]

I agree with this - I think there's kind of a separation between what goes on with the top level guys who have a Pultec or GML or whatever that they leave alone and run everything through, and the kid in their bedroom thinking that their plugin emulation of it ISN'T QUITE good enough and is making their work sound inferior to the guy they look up to.

The reality is those big names would produce exactly the same quality of work with or without any particular gear. If their Pultec is adding 3dB of top end, and the tracks that come in are already too bright, they're just going to compensate by EQing the tracks differently. Most of the time they're barely even thinking about gear because it largely gets in the way of actually doing the work. It's fun to own and can be a cool inspiring thing to have in the studio. A random ass NAM file doesn't really scratch that itch and just leads whoever's using them to have something else to doubt in their mixing chain.

But yeah, there are for sure a lot of people who are willing to throw money at all kinds of stuff in the hope that its the silver bullet and the solution to what's been holding them back. There are still COUNTLESS new plugins released of the same old gear - Pultecs, 1176's, LA2A's, Fairchilds, dbx160 etc, all as if whatever is out there isn't QUITE good enough and that this new one has some kind of minute difference that will suddenly translate into better results. It's wishful thinking but people are insecure and will think that the differences matter.

Meanwhile you have the top level guys out there mixing 3+ songs a day with old ass waves plugins not worrying about aliasing or oversampling or whether their SSL eq is more accurate than another one. They're just doing the work, getting it approved and moving on to the next one as quickly and efficiently as possible.

I have WAY too many plugins, but the reason for using different ones is for inspiration and fun. I could mix just fine with stock plugins but using 3rd party stuff keeps it interesting and inspiring. Going through a list of presets (be it in something like ControlHub) or NAM files doesn't really allow you to get into the song, its just swinging in the dark and won't really get very far.

I just caught a CLA clip and he has the same two plugins on every track pretty much. You make the sauce, you don’t buy the sauce.

 

[EOengineer]

I just caught a CLA clip and he has the same two plugins on every track pretty much. You make the sauce, you don’t buy the sauce.

What was he using? CLA 76 and ??

I will suggest that there’s a pretty good chance that if CLA is mixing your tracks you probably tracked in an environment where there’s plenty of mojo gear being used on the way in.

 

[ian_dissonance]

What was he using? CLA 76 and ??

I will suggest that there’s a pretty good chance that if CLA is mixing your tracks you probably tracked in an environment where there’s plenty of mojo gear being used on the way in.

It was the 76 and the CLA mixhub. And of course his tracks are all cleaned up before he touches them, but that even further shows the point that buying more plugins isn’t going to solve a problem. The thing I was watching everything but the guitars was recorded remotely by the musicians except vocals, and there’s a ton of samples for the drums, etc. But still, no magic bullets, just an ear and a couple of plugins.

 

[Ed DeGenaro]

I find it interesting how spending money on hardware systems an Axe3 for its amazing verb, or spending money on preset packs is okay around here but spending money is foolish.
It’s just another new toy for most. So what?

Plus at this point in the game there will be a whole lotta more music produced that won’t make a penny.

Again so what.

I started this thread to see if anyone had any experience with it. Not why something doesn’t make sense according to guys no having used it.

 

[EOengineer]

I find it interesting how spending money on hardware systems an Axe3 for its amazing verb, or spending money on preset packs is okay around here but spending money is foolish.
It’s just another new toy for most. So what?

Plus at this point in the game there will be a whole lotta more music produced that won’t make a penny.

Again so what.

I think we inadvertently veered into a critique of the larger tech world’s obsession with AI and torched the intent of this thread in the process.

I don’t think anybody is intentionally poo-pooing your purchase. I’m interested to see how it works out for you and how it matures. Capture tech can obviously be done well.

 

[Ed DeGenaro]

I think we inadvertently veered into a critique of the larger tech world’s obsession with AI and torched the intent of this thread in the process.

I don’t think anybody is intentionally poo-pooing your purchase. I’m interested to see how it works out for you and how it matures. Capture tech can obviously be done well.

Oh I’m even good with folks thinking something is crap…after they actually try it.

I mean it really is a “the farmer doesn’t eat what he isn’t accustomed to” attitude when it comes to new gear.

 

[ian_dissonance]

Oh I’m even good with folks thinking something is crap…after they actually try it.

I mean it really is a “the farmer doesn’t eat what he isn’t accustomed to” attitude when it comes to new gear.

I love new gear! And new plugins! I just prefer modeling to profiling/capturing in both instances. How are you applying this to your DAW projects?