leider ist mein kommentar und der zugehörige thread nachdem ich das folgende als meine abschluss message hinterlassen habe und auch auf strikt falsche claims im video hingewiesen habe sofort entfernt worden, schade hat lange gedauert zu schreiben...
Das war es:
As there isn’t much movement here anymore, and I also feel that I’m repeating myself in many comments, I decided to do one last comment here to wrap this conversation up.
It was very interesting and also valuable for me, but of course, at least partly, not very fun as well—mainly not because of the criticism toward my views and persona but because I think it’s a pity that there are such strong opinions and projections on something that should be mainly fun and connecting rather than dividing. And of course, the most emotional aspects of the conversation have been related to things that are much bigger than my plugin or anything I do business-wise. As you can imagine, spending a vast majority of my time developing and researching Artificial Intelligence—and based on the assumption that I’m not a horrible person who wants to trick people into giving me money, and that I’m actually focused on creating value for others (which, of course, also leads to money and is not charity work)—I have strong concerns regarding a lot of different AI topics myself, and the critical conversation is important, valid, and not one-sided.
Regarding the review: up to this moment, I do not know if the idea was, from the beginning, to put my plugin in the worst light possible to cater to an audience that clearly has, in a vast majority, negative views on it’s concept anyway or more likely, as Dan seems to be a nice guy, the review ended up completely ignoring what the concept of the plugin is—simply because my conceptual ideas and his views could not be further apart.
The following will involve a critique of his review, focusing solely on facts and evidence. If there are comments regarding the next points that are really not subjective (and don’t get me wrong, subjective critique is also valid), I hope there will be proper knowledge and research backing those comments up. Some stuff is very technical, and some stuff is simply things that have been (I assume unintentionally) left out of the video regarding the plugin.
Just a few things first, before I go into my notes on the video and what Dan is claiming, that I want to mention because I have seen comments often that seem like they have been made without actually checking what the concept and functionality of the plugin is I will do so without including anything about my subjective opinion.
1. The plugin is for beginners onlyI understand that my F1 driver comparison could have led to that assumption; I tried to find something flattering because I don’t feel like becoming hostile. But I did not mean that it caters only to beginners—it’s for those who look for a new/different workflow, no matter their experience. I use it myself—I would not consider myself a beginner. If the concept (and also the marketing) does not resonate with you, it’s not for you. You have established a workflow you are comfortable with—completely fine—you already have what you need.
2. I read buzzwords like “black-box” and all the related stuffThe plugin sets parameters for you after analyzing—it obeys a lot of rules that have been established over decades and that turned out to be effective from a technical point of view (gain staging, compressor settings for different signal types, for which signal types different effects are commonly beneficial, and so on).After analyzing, the plugin fully breaks down why things have been set the way they did, with explanation.Every parameter set by the plugin is tweakable by the user. The chatbot has this information (analysis and reaction), and the user can specifically ask why and how. For example, “I think it sounds too bright—what should I do?” is the opposite of a black box—there’s full emphasis on understanding and learning if wanted.
Now the actual video feedback:
The analysis of the signal is probably one of the most integral parts of the whole concept—skipping it already defeats the purpose of the entire approach. Everything that would happen next is purposeless, as every single element in the processing chain is built to catch up on what the result of the analysis is, primarily on a gain-staging perspective but also on many others.
The resonance EQ – which turned into reducing the whole plugin to “IT CAN NOT BOOOOST?”The only point of that module is reducing UNWANTED stuff before further processing. That’s the only thing it should do—nothing less, nothing more.That the plugin identifies potential problem frequencies based on the signal for the user to look at is not part of the video (and also not relevant, as the plugin has no idea because the analysis was skipped). I would say the frequency suggestion—based on an “unusual” buildup of energy at unexpected frequencies—is a pretty cool feature.Ignoring the purpose of this section and focusing on why it cannot do the opposite of its intention is pointless. The whole part about resonances and how they are integral to music is pointless, because the purpose is only to get rid of UNWANTED resonances, not those that are integral and positive. Dynamics are as integral to music—I would not say taming them with a compressor is evil.Making the resonance EQ able to push is one line of code less—the limitation is set with a purpose. And maybe it came across as if this was something random I did—it’s not.
Let’s get nerdy:Boosting and cutting frequencies in an EQ is NOT the same.A -3 dB cut lowers band-limited energy without raising absolute peaks, so the signal’s crest factor and converter margin are unchanged; a +3 dB boost reduces available headroom by the same amount and can push fixed-point DSP into partial clipping without even noticing it (inter-sample peaks).Cuts attenuate both program and broadband noise in the selected band, leaving the overall SNR intact or improved; boosts increase the RMS level of the filter output and therefore amplify pre-existing hiss by the boost factor.
Most importantly,Phase & impulse behavior: For a given Q, the magnitude of phase rotation Δφ(ω) is proportional to the gain factor G; with |G| < 1 (cut), the rotation is smaller, giving shorter group-delay tails and less temporal smearing than the same filter run at |G| > 1 (boost). A boost makes that phase coloration and the associated ringing more obvious because you’ve also made that band louder, whereas a cut naturally masks it.Non-linear stress: Lower post-filter amplitude from a cut keeps subsequent stages in a more linear operating region, minimizing harmonic and intermodulation products; higher amplitude from a boost increases drive and distortion proportionally.
In short, cutting frequencies—if not heavily overdone—will basically never introduce new problems; boosting does (also for linear phase EQs). The harm caused is probably not really noticeable, but it defeats the purpose of this section:Cleanup stuff that shouldn’t affect further processing—only if needed.It’s surprising to me that this is so polarizing, as I did not invent subtractive-only EQing—neither as a technique nor as a limitation. Feel free to look up the following beautiful devices: Urei 539 “Room Equalizer,” Altec-Lansing 529A “Acousta-Voicette,” White Instruments 4200 / 4220 series, Klark Teknik DN301.
A fair and valid review of the section would have been: how good is it at identifying potential problem frequencies, and how good/easy is the job it’s made for? It’s a dynamic EQ because this makes a lot of sense if unwanted resonances only appear from time to time. The “quality” of the signal processing of the EQ behind it is industry standard (it’s not witchcraft DSP).
Next

eak-Catcher—yes, it’s simply there to catch occasional peaks, so stuff that shouldn’t trigger the further processing. One thing I need to correct is the claim that the bot doesn’t know it applies gain based on the reduction. Also, if this doesn’t make sense for Dan, here is the explanation: The highest peaks are biasing the overall peak of the signal—if I get rid of them by squashing them, the highest peaks have been moved to a new baseline, so I’ve created some headroom that I can now use again for the purpose of having the average peak hit the sweet spot of the DSP algorithm. The actual compressor has a limit of 10ms attack, so the peak catcher and the actual compressor are separated in the signal flow. Turning the threshold of the peak catcher down until everything is squashed is a pointless demonstration against the purpose of the section once more. Yes, this sounds bad, agreed—this would sound bad with any fast compressor on the market, no matter the brand. You showcased a hard limiting of a total of -10 dB lower than the highest peak with an ultra-fast attack and release and an ultra-high ratio and hard knee. Measuring the highest peak (it’s a bit more advanced than a max peak readout—I made an algorithm that analyzes hundreds of highest peaks, cleans it up based on how often they occur, and so on—it’s more of a hybrid) rather than pure RMS is in this case much more efficient, as we are controlling the highest peak in multiple stages and the predictability is much higher.
This leads us to floating point and the attempt to make the bot look stupid, and I wish the answer would have been the following:Yes, Dan—your DAW is floating point, which doesn’t make headroom a problem in isolation. What you are missing, though, is that even though plugins technically can handle floating point, any DSP algorithm that adds coloration (and this includes all plugins that you like for the “sound”) is creating this coloration by using non-linear transfer functions—which are, as the name suggests, non-linear. The signal after passing those functions is NOT restorable anymore.Additionally, most DSP algorithms are calibrated to a -12 dB input—this is not random; this is based on the analog domain, which we all love so much because it gives some character to our perfect computer sounds, even when emulated. The reason behind that is that -12 dB is close to 0 VU, which is the sweet spot in the analog domain (often even less, like -18). The linear headroom is very important, especially as linear functions stack up in the processing. So relying on floating point and not considering this gently makes the quality of your signal worse and worse. This is also why auto-gaining to -12 dB at every stage is such an important thing in the plugin.
Next Transient Designer:There is a HoldTime slider on the far left, which is set to 1/16, making it not allow triggering for one 16th note after a transient is triggered to avoid double triggering. This gives the ability to only trigger half or quarter notes, for example. With an unquantized drum loop, this leads to unexpected behavior. This is really cool for programmed drums but not ideal for other cases, which is something I feel I should address, as it is an actual limitation that I wasn’t focusing on in development because the field of music I primarily do—and that most of my users do—is more programmed. But it is an oversight I want to address.
Next, you raise the question “What if I don’t like the stuff this is trained on?” and hover over Profile Designer without clicking it, assuming: “If you can do your own profiles, you won’t need this plugin anyway.” Truth is: creating a profile is as simple as entering a text description of what the signal is, adding a reference audio file, and then waiting one second. A feature I’m pretty proud of, factually ignored, to make a point that caters to the conclusion.
Summing it up:If we agree that the purpose of a product review is to test the product based on what its purpose is, then this is not a review; it’s definitely a funny piece of YouTube content that ignores the purpose and concept constantly and highlights only things that contradict it. I think I have added enough neutral examples of that above.
A proper treatment would have been to take a couple of unmixed stems, let the plugin do its job, and then talk about the result from a technical standpoint and also a subjective one (do I like the aesthetic, and how easy is it to tweak it to get there?). Taking time for features that might not be so useful for the reviewer but could be valuable for others would probably have been the cherry on top.
Subjective stuff:Of course, it’s not fun to watch for me—I put a lot of work and thought into it—but that’s not why I’m taking so much time here. I do not believe that it was the intention, but it is a hook for a much bigger topic, leading to a lot of dissatisfaction, anger, and, in the end, creating negativity in a space that should be fun, creative, and fulfilling. I hear the concerns regarding AI—I have many myself. But the idea that it will destroy human music is the least of them. Statistically, lowering the entry barrier to music production has only led to more humans making more music and earning more from it as well. Think SoundBetter, Splice…In a world where everything becomes more algorithmic and generated, the desire for real human emotion translated through human music can only grow. Maybe I’m naive. But I’d rather see the future that way, and I have a bit of knowledge to back it up. No matter if you use AI or not, make music that makes you happy in whatever way works for you. If you love it, somebody else will too. If you do not like my approach to things—just ignore it—it doesn’t take anything away from you if somebody else has fun with it.
One last thing before I’m out: The idea that my plugin can do harm to amateurs, and all this type of argumentation, assumes that a beginner is an idiot who doesn’t know how things should sound (whatever that means). I think this is not only completely untrue but also simply gatekeeping. It makes beginners feel dumb, discourages them, and takes away some of their fun without benefitting anyone. (I read “I’m ashamed to mention the DAW I’m using” in the comments.) WTF? I know people who can give better mixing feedback without even knowing what a compressor is than people who own 10 analog compressors. It’s mostly not the vision—it’s the how-to that is missing. Or is this what you are afraid of—that the ones with better ideas suddenly have a lower entry point to make them real?