the case for fewer controls (long version)

the year of the macro

a major plugin company shipped a desktop hardware controller in April 2026 with a tagline about “transforming sound instantly.” the four large knobs on the controller are not assigned to anything fixed. each one is a macro, it controls some weighted combination of the underlying parameters in whichever signal chain you loaded. the marketing language called it the future of mixing.

a different major plugin company sells a compressor with eight broadband algorithms, three character modes, a six-band sidechain EQ, knee, ratio, threshold, attack, release, lookahead, oversampling, range, hold, dry/wet, output, an output peak limiter, and a stereo link toggle. that plugin costs $199.

a third company ships a free OTT-style multiband compressor with an X/Y pad replacing the multi-band controls. drag horizontally for tone, drag vertically for depth. that is the entire interface.

three plugins, three answers to the same question, how many controls should a compressor have?, and the answers range from 2 to 17 to a macro-knob abstraction layered on top of an unknowable count. (this is also where producers find themselves when they open Plugin Boutique in 2026 and there are 9,000 plugins to pick from.)

i think the field has the question wrong. the real question is not “how many controls” but “which controls are load-bearing?”, and the second question rules out most of the answers to the first.

why “more” stopped being the answer

audio plugins inherited their interface conventions from analog hardware. a hardware compressor had as many knobs as the chassis could fit. each knob mapped to a circuit. removing a knob meant removing a feature. the constraint was physical, and the result was that engineers learned to mix on plugins shaped like the boxes their mentors had used.

two things have changed in the last decade. the first is that DAW screens are not chassis. there is no physical reason a software compressor needs a lookahead toggle and an oversampling toggle and a stereo-link toggle and a dry/wet knob and an output trim and a 6-band sidechain EQ. the second is that producers are not ten-year apprentices anymore. the modal user is somewhere between 22 and 38, mixes their own work, and learned audio on YouTube. for that user, every additional control is a decision they did not ask to make.

decisions in mixing are expensive. there is well-established research on choice overload, the more options a user faces, the lower the quality of the decision and the slower the time to commit.[^1] in audio, this shows up as the producer who spends 40 minutes on a vocal compressor preset and still does not know which one is right. in software design, this shows up as the famous Pareto-style usage distribution: in most rich plugins, 80% of users touch the same 5-7 controls and never go near the others.

if 80% of users touch the same 5-7 controls, the rational design is to ship a plugin with those 5-7 controls and put serious work into making each one a real decision. that is also the position major commercial plugins do not take, because feature parity is how plugin reviews score.

(this is the part most tutorials skip. nobody at iZotope wants to ship a 5-knob plugin because the review headline writes itself: “missing critical features at this price point.” the structure of the market punishes restraint.)

the load-bearing test

a control is load-bearing if the answer to all three of these questions is yes:

1. does this control change the audible result in a way the producer can predict? if changing the knob produces a result the user could not have anticipated, the control is making decisions the user is not.
2. would removing it cause a real workflow problem? if removing the control means a meaningful job becomes impossible, the control earns its place. if removing it just means “you would have to use a different preset,” the control is decoration.
3. does it map to a single, namable decision in the producer’s head? “how much compression” is a decision. “knee width” is decoration unless the user can articulate when soft vs hard knee is the right choice for their material.

most plugin parameters fail at least one of these tests. lookahead. oversampling. stereo-link percentage. range cap. dry/wet on a non-parallel-compression plugin. these are settings, not decisions, they should default to sensible values and stay out of the user’s way.

what 5 knobs does to a plugin

every plugin in the KERN catalog has exactly 5 knobs. this is not a marketing constraint. it is the result of repeatedly applying the load-bearing test until each plugin’s control set stabilised, and 5 turned out to be the natural ceiling for each of the four DSP problems we ship.

KERN SMOOTH, resonance suppression. five controls: depth, sharpness, selectivity, tilt, mix. each one maps to a single decision. how aggressive should the suppression be? how surgical should the response be? which frequencies should the plugin focus on? how does the suppression bias across the spectrum? how much of the dry signal stays in?

KERN WARM, saturation. five controls: drive, body, character (a mode pick rather than a knob, but conceptually a knob), tilt, mix. drive sets harmonic content. body sets the spectral region the saturation lives in. character picks the topology (tape vs tube vs transformer). tilt biases. mix blends.

KERN WIDE, stereo expansion. five controls: amount, focus, mid level, side level, mix. amount sets the expansion. focus picks where in the spectrum it acts. mid/side levels manage the stereo balance directly. mix blends.

KERN PUSH, compression. four knobs and two toggles: amount, speed, focus, mix, plus DELTA and M/S. character is a mode pick (SETTLE / TIGHT / DRIFT). amount sets compression depth. speed scales attack and release together. focus picks the spectral target. mix is dry/wet.

KERN CHECK, mono compatibility. one mode pick (display type), one routing toggle (M/S), one ZOOM action. analysis-only, zero gain stage, no audible processing.

each of these plugins could have shipped with twice as many controls. the load-bearing test would not have stopped me from shipping fifteen knobs each. listening tests did. every additional control past 5 made the plugin harder to use without making the result audibly better. that is a feature decision, not a marketing decision.

(the brand law that says “max 5 knobs” was not the cause; it was the result. the cause was running the test honestly. the law is the artifact.)

the AI question

the obvious counterargument: AI plugins remove decisions entirely. why not just hand the producer a one-knob “make this sound better” interface backed by a model trained on a million masters?

a one-knob AI plugin is not the opposite of a 17-knob plugin. it is the same plugin with the controls hidden. the parameters still exist; the user just does not get to see them or override them. the actual question is who makes the decision, the user, by turning a knob, or the model, by predicting what the user wanted.

both approaches work for some users on some material. but they are not interchangeable. an AI plugin gives you a result, fast, polished, often reasonable. a tool with explicit controls gives you a technique, slower, more variable, but reusable across projects.

the difference compounds. a producer who got handed a compressed vocal does not know what the compressor decided. the next song is a different problem; they have to ask the AI again. a producer who turned an attack knob from 30 ms to 5 ms knows what they did and will know to do it again next time. one of these scales with the producer’s experience; the other does not.

KERN plugins are built for the producer who wants the technique, not the result. that is a position, not a universal claim. for someone who wants the AI result, those plugins exist and some of them are excellent, Sonible’s smart series, iZotope’s Master Assistant, Softube Flow’s signal-chain macros. they belong in the toolbox of producers who use them as designed. they do not belong in mine, because i am a producer who wants the technique.

(if “fewer controls, AI behind the scenes” is the future of plugin design, KERN is not in that future. that is fine. there will be a future for plugins where the user wants to keep the decision, and that is the future i am building for.)

what gets gained when you cut

a plugin with 5 controls instead of 17 is not just smaller. three things actually get better:

first, the producer gets faster. every control on a panel is a decision that has to be made or actively skipped. five fewer controls means five fewer decisions per session. multiply that by 30 plugin instances on a mix and the time saved is real.

second, the developer gets to spend the budget on the controls that remain. a plugin with 5 controls means 5 controls each have months of tuning, A/B testing, and listening-test work behind them. a plugin with 17 controls divides the same budget across 17. KERN PUSH‘s SETTLE character had four months of tuning between v1.2 and v1.4, almost all of it on a single hidden coupling constant inside the gain-reduction circuit, plus the spectral tilt curve. that depth of tuning would not have been possible if the budget had to cover 12 more parameters too.

third, the plugin gets cheaper to ship. fewer controls means fewer code paths, fewer tests, fewer edge cases, fewer support emails. KERN’s $29 price is sustainable partly because the plugins do not have a shadow QA cost from features no one uses. (the cynical version: plugins shipped with 17 controls because their developers were optimizing for a review score, not for a producer outcome. price tracks that.)

the cost of cutting

restraint is not free.

a 5-knob plugin cannot do everything a 17-knob plugin can do. someone who needs lookahead-style transient detection on a master limiter is not going to get it from a glue compressor with attack and release as the only timing controls. someone who wants to A/B-compare 14 different broadband compressor algorithms on the same material has to load 14 plugins instead of one. these are real workflows, and the 5-knob plugin loses them.

the calculus is whether the workflows lost matter to the producer in front of the plugin. for the modal KERN user, independent producer working in Ableton on their own material, primarily mixing pop / hip-hop / electronic / hybrid material, the workflows lost are uncommon enough that the trade pays back. for a mastering engineer working on classical recordings with extreme dynamic-range needs, the trade does not pay back. that engineer should buy the FabFilter Pro-C 3, and i would tell them so.

i do not think 5-knob plugins are universally correct. i think they are correct for a specific population of users with a specific kind of work. when i misidentify the population, the plugin lands flat. that has happened. it will happen again. the position is not “fewer is always better”; it is “fewer is better for the user we are building for, and we have committed to that user.”

the broader picture

every plugin company has implicitly answered the question “what kind of producer are you serving?” with their control count. fifteen knobs says we serve the engineer who wants every option. one X/Y pad says we serve the producer who wants speed above all. four macros mapped to a hidden chain says we serve the producer who wants curated results. five honest knobs says we serve the producer who wants to make the decision themselves and have it stick.

none of these positions is wrong. they are different bets about who the user is, and the market has space for several. the part i find dishonest is the plugin that ships with 17 controls and a marketing claim about being “intuitive”, those words mean opposite things. choose one or the other.

KERN’s bet is that the producer who wants to make the decision themselves is a real and underserved population in 2026, and that they have grown frustrated with the alternative, plugins that hide the work, AI that picks the settings, macro chains that flatten the technique into a knob. the bet might be wrong. so far the people i hear from say it is not.

(if you have read this far, you are probably the user this article is about. that is also fine.)

references

a note from the developer

the first version of KERN WARM had 9 knobs. drive, body, tilt, character, mix, output, dry-side level, mid level, side level. the design felt right on paper. then i sent it to four producers for feedback and watched them mix with it for a week. nobody touched the dry-side level. nobody touched the output knob. one producer touched the mid and side levels but only to set them once at the start of the session and never again.

i cut four knobs out of WARM the next morning. the plugin shipped with 5. nothing was lost in the process, every removed knob got folded into a sensible default, and the things people actually mixed with were now front and centre with more room to breathe. the panel got cleaner. the support emails dropped. the average time a producer spent on a saturation pass dropped by maybe 30%. i have no hard data on that last one; it was an impression from the testers.

every plugin in the catalog went through the same process. it was not a brand exercise. it was repeated rounds of watching producers actually use the plugins and cutting whatever they did not touch. the result is a 5-knob convention that landed not because i decided in advance to ship 5-knob plugins, but because 5 was the count where the cuts stopped feeling productive.

i do not believe my approach is the only one. plugins like FabFilter’s Pro-C 3 and Sonible’s smart:comp 3 are excellent at what they do, and they are not built for my user. i build for someone specific: an independent producer in Ableton who wants the decision, can hear the difference, and is bored of plugins that try to hide the work behind a macro or an AI. if that is you, the catalog is for you. if it is not, there are excellent plugins that do not look anything like KERN’s, and i hope you find them.

if you have a take on plugin design that took years to arrive at, especially if it disagrees with what i wrote here, send it. jonas@kernaudio.io. brand voice is built by argument, not consensus.