Wonder if anyone saw this? https://www.reddit.com/r/audioengineeri ... aturation/
There's also a youtube clip:
These plugins got such high praise, even if what is claimed is true, does it bother you at all that "Omega-TWK seems to be literally just a tanh() function (or something similar like x/(1+abs(x)) which is staggering to me, as its selling for $29USD!!"
Kush plugins "reverse engineering"
-
- KVRAF
- 6390 posts since 8 Jun, 2009
I'd be cautious of anyone who claims to match a plugin just with static waveforms. Maybe the Kush one is that simple. But this guy seems to have convinced himself he doesn't need to look at things like transient response to see whether he's matched something or not.
If his stuff is free, there's not a big risk in finding out, but does his version null with the Omega under anything that isn't a constant sine wave?
If his stuff is free, there's not a big risk in finding out, but does his version null with the Omega under anything that isn't a constant sine wave?
-
- KVRian
- 888 posts since 22 Jan, 2022
This is very informative, and what a lot of folks don't understand about emulation plugins is that many of them are just a basic repackaging of a waveshaper behind a sexy skeuomorphic UI.
Some companies take this further by combining waveshaping and convolution technology (Acustica). Others like Kirchhoff are using circuit modeling tech to re-duplicate the nuance of each component and flow within the circuit.
These approaches come at a cost. Convolution can create large plugin sizes, non-continuous controls, and laggy responses. It also doesn't fully capture non-linearities, so often has to be mixed with other tech to round out the emulation. Circuit modeling is great, but typically comes at a very high CPU cost. There's no free lunch.
The best approach to creating emulations would probably be some mix of convolution, circuit modeling, and high oversampling rates. But a lot of novel optimization techniques will need to be created before someone can reasonably expect to drop one or more of these hybrid emulators on every track and buss to mimic a full analog signal path.
Would all this be worth it? At this point there are tons of successful productions that never saw an analog circuit outside of the initial vocal tracking.
Some companies take this further by combining waveshaping and convolution technology (Acustica). Others like Kirchhoff are using circuit modeling tech to re-duplicate the nuance of each component and flow within the circuit.
These approaches come at a cost. Convolution can create large plugin sizes, non-continuous controls, and laggy responses. It also doesn't fully capture non-linearities, so often has to be mixed with other tech to round out the emulation. Circuit modeling is great, but typically comes at a very high CPU cost. There's no free lunch.
The best approach to creating emulations would probably be some mix of convolution, circuit modeling, and high oversampling rates. But a lot of novel optimization techniques will need to be created before someone can reasonably expect to drop one or more of these hybrid emulators on every track and buss to mimic a full analog signal path.
Would all this be worth it? At this point there are tons of successful productions that never saw an analog circuit outside of the initial vocal tracking.
Last edited by billinder33 on Sun Feb 15, 2026 7:26 pm, edited 1 time in total.
-
- KVRist
- 484 posts since 8 May, 2007
Yes, of course they are. For one thing, tanh(x) is odd-symmetric about the origin as is x/(1+abs(x)). Real behavior is often asymmetric.lobanov wrote: Sun Feb 15, 2026 3:31 pm Do you think other developers offer you something more sofisticated? I am not sure.
See for example, Will Pirkle's excellent discussion of tube distortion:
https://willpirkle.com/special/Addendum ... e_v1.0.pdf
I myself have made us of quasti-static simulations in SPICE which produce asymmetric waveshapers as well as Will's dynamic grid-conduction modelling in my own amp sims. Revalver includes SPICE circuit analysis, albeit curtailed due to CPU performance issues, but nevertheless has an acknowledgement of SPICE 3 (If I recall the version correctly) from Berkeley. I would guess that it's some sort of partitioned MNA solution plus at least some N-R for nonlinear devices.
I have not watched the cited video yet, but I don't need to to write the following: Any claim that developers have not moved beyond tanh and the like is absolute nonsense.
-
- KVRAF
- 11375 posts since 3 Feb, 2003 from Finland, Espoo
Waveshaping is just a tool, like any others. How you use that tool and manipulate the results (pre/post filtering, manipulating the symmetry etc) is where the secret sauce lies. That's what you pay for, the ear/taste of the one who used those tools to manipulate the output. And if you don't want to pay anybody you can always create your own. It's not rocket science.. and in the end, it's all subjective. That subjective knowledge and taste is what you ultimately pay for. Not the building blocks behind the scene.
However, things do get a bit murky when companies claim "exact models" of known hardware. That's when I think it's okay in this day and age to call them out. To be fair, if a hardware model only produces simplistic specific harmonics (the ones that actually make a difference) and are slightly pre/post filtered, and this satisfies and matches the output of the device, then I think it's fine to use simple tools to get the job done. The proof is always in the pudding, aka listening tests. Doesn't matter what technique was used if nobody can tell the difference.. but that is usually where simple things fall apart, people CAN tell a difference.
However, things do get a bit murky when companies claim "exact models" of known hardware. That's when I think it's okay in this day and age to call them out. To be fair, if a hardware model only produces simplistic specific harmonics (the ones that actually make a difference) and are slightly pre/post filtered, and this satisfies and matches the output of the device, then I think it's fine to use simple tools to get the job done. The proof is always in the pudding, aka listening tests. Doesn't matter what technique was used if nobody can tell the difference.. but that is usually where simple things fall apart, people CAN tell a difference.
"Wisdom is wisdom, regardless of the idiot who said it." -an idiot
"They don't ban hate speech; they ban speech they hate." -an oracle
"They don't ban hate speech; they ban speech they hate." -an oracle
-
- KVRian
- 871 posts since 20 Jun, 2010
If that guy is so certain he reverse engineered it, why not use some real world examples instead of just sine waves? I don't know (or care) if the omegas are just waveshapers, but imo this video doesn't prove anything.
-
- KVRian
- 1223 posts since 11 Aug, 2004 from France
This video is the very definition of ultracrepidarianism. What is shown here gives no information about what is actually in the original plug-in. And just sending sine waves into a plug-in or any piece of hardware would really do the worst job ever for understanding anything about its behaviour. That's the reason why tools like PluginDoctor for example can do much more sophisticated tests than that.
About waveshapers and specifically the operation tanh, as someone said before, they are just some of the DSP tools available to create FXs which add nonlinearities, from the simplest to the ones doing proper virtual analog modeling, even where asymmetry is involved, since you can have it with a tanh, depending on what you put in the argument or the functions outside
I have seen a lot of videos like this on Youtube recently, with fake experts claiming they are able to find out when an audio company is "scamming" its users, and sometimes trying to sell you something they came up with thanks to their amazing discoveries
I'm not saying that everything is perfect on the audio companies side, actually it's not, but it's a bit pathetic to try to do some business with fake and wrong bad buzz like that, when you could also be honest with your audience, and try to cultivate one skill or two to make something useful and fair.
About waveshapers and specifically the operation tanh, as someone said before, they are just some of the DSP tools available to create FXs which add nonlinearities, from the simplest to the ones doing proper virtual analog modeling, even where asymmetry is involved, since you can have it with a tanh, depending on what you put in the argument or the functions outside
I have seen a lot of videos like this on Youtube recently, with fake experts claiming they are able to find out when an audio company is "scamming" its users, and sometimes trying to sell you something they came up with thanks to their amazing discoveries
I'm not saying that everything is perfect on the audio companies side, actually it's not, but it's a bit pathetic to try to do some business with fake and wrong bad buzz like that, when you could also be honest with your audience, and try to cultivate one skill or two to make something useful and fair.-
- KVRian
- 698 posts since 9 Dec, 2021
It's easy to tell if a plugin is not actually doing proper emulation just by looking at the CPU usage, this stuff is not light at all if done properly. People like Andy from Cytomic has to spend a decade just to optimize the tech, or Relab who rawdogged it with the 176. 95% of the "emulations" on the market are just clever tuned dynamic waveshapers with some EQ/filters + compression, it can sound really good if the dev is skilled, but it's not real circuit model and will show it's weakness when going for extreme settings or configurations that the dev is unaware of (if the dev doesnt know about 1176 all button, then the plugin wont have it, whereas a full simulation will capture everything)
Nothing wrong with using waveshapers btw, the only problem is the marketing: "every component modelled, end to end..." shut the hell up.
Nothing wrong with using waveshapers btw, the only problem is the marketing: "every component modelled, end to end..." shut the hell up.
Last edited by jtsterays on Mon Feb 16, 2026 7:04 pm, edited 1 time in total.
-
- KVRian
- 888 posts since 22 Jan, 2022
This guy never said Kush is a scam (he actually said they make nice plugins), and all the M4L devices listed on his site are donation only, free if you want. So he's not exactly slamming other's work in an effort to hard-sell his own wares. IMO he's being very reasonable and transparent about his comparison methods.Ivan_C wrote: Mon Feb 16, 2026 12:52 pm This video is the very definition of ultracrepidarianism. What is shown here gives no information about what is actually in the original plug-in. And just sending sine waves into a plug-in or any piece of hardware would really do the worst job ever for understanding anything about its behaviour. That's the reason why tools like PluginDoctor for example can do much more sophisticated tests than that.
About waveshapers and specifically the operation tanh, as someone said before, they are just some of the DSP tools available to create FXs which add nonlinearities, from the simplest to the ones doing proper virtual analog modeling, even where asymmetry is involved, since you can have it with a tanh, depending on what you put in the argument or the functions outside
I have seen a lot of videos like this on Youtube recently, with fake experts claiming they are able to find out when an audio company is "scamming" its users, and sometimes trying to sell you something they came up with thanks to their amazing discoveries
Kush claims their waveshaping has been altered using Transient Harmonic Coupling (THC - I assume this is a term they made up to be witty. Kush uses THC, get it! har har) to alter the response based on speed and slope of the input signal to affect transients, likely by placing a slew rate algo in front of the waveshaper.
If you are just running a steady state sine wave against both sets of code (Osprey's and Kush's) then you are only ascertaining the harmonic content at steady 'voltage' but not the slew rate. Slew rate can make a big difference and provide that 'dark/thick' sound that Neve pres are known for. I have Darkplace Studios Sloth which is an excellent slew rate tool, and can make a huge difference in how 'analog' something sounds. So if I'm to decode Kush's flowery markitecture into their real world implementation it's Input > Slew > Waveshaper > Output. More then just Tanh, but not exactly groundbreaking stuff.
For someone who just wants a simple one-knob implementation of slew+waveshaper curated to specifically emulate a Neve, Omega is a pretty good deal at $20. For more control, Sloth is 20€ and you can find free (and paid) waveshapers everywhere. For most people who just want "sounds like Neve" this probably isn't worth the hassle, and IMO most of this stuff is way overblown in terms of real world importance anyhow.
The one thing I wish Osprey had done is run a null test between his device and the Kush, both on the sine wave and also on some program material. Maybe an Kush+Ableton owner would like to do this and report back.
Last edited by billinder33 on Tue Feb 17, 2026 12:05 am, edited 1 time in total.
-
- KVRian
- 1156 posts since 2 Oct, 2021
This!billinder33 wrote: Mon Feb 16, 2026 4:58 pm The one thing I wish Osprey had done is run a null test between his device and the Kush, both on the sine wave and also on some program material. Maybe an Kush+Ableton owner would like to do this and report back.
ABX is enemy to GAS
-
- KVRian
- 829 posts since 7 Oct, 2005
Oh, thank you for the answer. Very informative. Yes, it is not fair to tell that all developers offer simplistic, primitive algos.DaveClark wrote: Sun Feb 15, 2026 6:04 pmYes, of course they are. For one thing, tanh(x) is odd-symmetric about the origin as is x/(1+abs(x)). Real behavior is often asymmetric.lobanov wrote: Sun Feb 15, 2026 3:31 pm Do you think other developers offer you something more sofisticated? I am not sure.
See for example, Will Pirkle's excellent discussion of tube distortion:
https://willpirkle.com/special/Addendum ... e_v1.0.pdf
I myself have made us of quasti-static simulations in SPICE which produce asymmetric waveshapers as well as Will's dynamic grid-conduction modelling in my own amp sims. Revalver includes SPICE circuit analysis, albeit curtailed due to CPU performance issues, but nevertheless has an acknowledgement of SPICE 3 (If I recall the version correctly) from Berkeley. I would guess that it's some sort of partitioned MNA solution plus at least some N-R for nonlinear devices.
I have not watched the cited video yet, but I don't need to to write the following: Any claim that developers have not moved beyond tanh and the like is absolute nonsense.