KERN WARM vs puffer:fish: two approaches to saturation

why this comparison matters

you want to add warmth to a vocal or glue to a drum bus. puffer:fish from Sonible is free and has three character modes. KERN WARM is $29 and also has three characters. Decapitator is $199 and is the industry standard.

both puffer:fish and KERN WARM offer three flavors of saturation. the similarities end at the surface. underneath, these two plugins take fundamentally different approaches to how they generate and control harmonics.

full disclosure: i make KERN WARM. this comparison gives puffer:fish the same honest treatment i give my own plugin. if puffer:fish is better for your use case, i will say so.

two approaches to saturation

saturation adds harmonics to audio. every saturation plugin does this by running your signal through a nonlinear function (a waveshaper). the question is what happens before and after that waveshaping step.

puffer:fish: conventional waveshaping with character selection

puffer:fish applies saturation using three character modes that shape the harmonic profile differently. the processing is applied uniformly across the frequency spectrum: what goes in gets shaped the same way regardless of whether it is a 100 Hz bass note or a 10 kHz cymbal.

this is how most saturation plugins work. it is effective and predictable. turn up the drive, get more harmonics.

KERN WARM: spectral-aware Chebyshev waveshaping

KERN WARM runs your audio through an FFT analysis pipeline before it reaches the waveshaper.[^1] the spectral engine (shared with KERN SMOOTH) analyzes the frequency content in 40 ERB bands and computes a per-frequency drive curve. this means the plugin knows which frequencies are present and how much energy they carry before applying any saturation.

the waveshaper itself uses Chebyshev polynomials with per-character coefficient profiles, processed through Baker’s 2nd-order antiderivative anti-aliasing (ADAA).[^2] this mathematical technique eliminates aliasing artifacts without the CPU cost of high oversampling ratios.

what puffer:fish does well

Sonible built puffer:fish as a free entry point to their smart: ecosystem, and it is better than most free saturation plugins.

three character modes. Tinyfin, Spikeskin, and Twitchgill each shape the harmonic profile differently, giving you tonal variety without needing separate plugins. the characters are musical and well-tuned, ranging from gentle warmth to aggressive bite.

Sonible engineering. this is the team behind smart:EQ 4 and smart:comp 3. puffer:fish benefits from their DSP expertise even at a zero price point.

clean interface. a single drive knob and a character selector. you can be adding saturation in seconds. the simplicity is genuine: there is nothing hidden.

zero cost, ecosystem gateway. free with no strings. if you like Sonible’s approach, their paid tools build on the same design philosophy.

where it falls short. puffer:fish has no mix knob (wet/dry blend) and no auto-compensation for loudness changes. this means you cannot parallel blend the saturated signal with the dry, and louder-is-better bias makes it harder to judge whether the saturation is actually improving your sound. without spectral analysis, it also applies the same saturation curve to every frequency. on a solo vocal or a simple stem, this works fine. on a full mix or a complex source with wide frequency content, the low-end energy dominates the waveshaper, generating intermodulation products in the midrange. you end up reaching for an EQ after the saturator to fix what the saturator introduced.

what KERN WARM does differently

KERN WARM is a hybrid spectral/time-domain saturation plugin. the spectral analysis pipeline (shared with KERN SMOOTH) controls the drive curve. the actual waveshaping happens in the time domain using Chebyshev polynomials.

spectral-aware drive shaping. the ERB filterbank analyzes your signal and computes a per-band drive multiplier. frequencies with less energy get proportionally more drive. frequencies already carrying heavy energy get less. this prevents low-end intermodulation while enhancing the presence range.

three analog character profiles. Tape (symmetric tanh soft clipper), Tube (bias-shifted asymmetric clipper generating even harmonics), and Transformer (quadratic asymmetry). these are not labels on the same curve. each character uses a different mathematical model based on how the real analog circuit behaves.[^3]

Baker ADAA. 2nd-order antiderivative anti-aliasing using Baker’s recursive integration identity for Chebyshev polynomials. combined with 2x oversampling, this achieves roughly 52 dB of alias suppression. without ADAA, digital saturation generates harmonics above the Nyquist frequency that fold back as inharmonic noise.[^2]

M/S routing. saturate the mid and side channels independently. add body to the center (vocals, bass) without affecting the sides (cymbals, room), or add width by saturating the sides more aggressively.

where it falls short. $29 is not free. no CLAP format. the spectral pipeline adds latency (~93ms). the three-character system, while technically deeper, requires more time to dial in than a simple drive knob. less established than Decapitator.

the comparison

comparing the characters

both puffer:fish and KERN WARM offer three saturation characters. the key difference is what those characters are actually doing under the hood.

puffer:fish’s three modes (Tinyfin, Spikeskin, Twitchgill) shape the transfer function differently, producing distinct harmonic balances. the characters are musical and cover a useful range from subtle to aggressive. notably, puffer:fish has no mix knob and no auto-compensation: what you hear is the fully saturated signal, and any loudness increase is left for you to manage.

KERN WARM’s three characters use fundamentally different waveshaping mathematics:

• Tape uses a symmetric tanh soft clipper, producing primarily odd harmonics (3rd, 5th, 7th). smooth, compressed, forgiving.
• Tube uses a bias-shifted asymmetric clipper, generating both even and odd harmonics (2nd + 3rd dominant). adds body and warmth.
• Transformer uses quadratic asymmetry (t-squared), producing strong even harmonics with a harder edge. adds presence and bite.

when to choose which

choose puffer:fish when:

• your budget is zero and you need saturation now
• you are working on individual stems where spectral complexity is low
• you want simple, quick operation
• you use Pro Tools (AAX support)
• you want to explore Sonible’s smart: ecosystem

choose KERN WARM when:

• you work with complex sources (full mixes, dense arrangements, layered vocals)
• you want saturation that adapts to the frequency content of your signal
• you care about alias suppression (clean harmonics without digital artifacts)
• you use M/S processing for independent mid/side saturation
• you want technically distinct analog character models, not just three variations of the same curve
• $29 is within your budget

choose Decapitator when:

• you need the industry standard that engineers recognize
• you want five character options with deeper tone shaping
• you need the highest oversampling for critical mastering work
• $199 is within your budget

frequently asked questions

frequently asked questions

is puffer:fish a good free saturation plugin?

puffer:fish is a well-made free saturation plugin from Sonible with three character modes (Tinyfin, Spikeskin, Twitchgill). it handles basic saturation tasks well, though it lacks a mix knob and auto-compensation. for spectral-aware saturation with anti-aliasing, psychoacoustic frequency shaping, and parallel blending, KERN WARM ($29) takes a different approach.

what is the difference between spectral-aware and conventional saturation?

conventional saturation applies the same waveshaping curve to the entire signal. spectral-aware saturation analyzes the frequency content first and adjusts how much saturation each frequency range receives. this prevents mud in the low end and harshness in the highs, problems that simple saturation introduces on complex material.

is KERN WARM worth $29 when puffer:fish is free?

it depends on what you need. puffer:fish handles basic saturation well. KERN WARM adds spectral-aware frequency shaping (the FFT pipeline controls where saturation is applied), Baker ADAA for alias-free processing, and three analog character profiles (Tape, Tube, Transformer) based on real analog circuit behavior. if you want saturation that responds to the spectral content of your audio, the $29 covers meaningful engineering.

what is Baker ADAA and why does it matter?

Baker ADAA (antiderivative anti-aliasing) is a mathematical technique that eliminates the digital artifacts (aliasing) that waveshaping creates. without it, saturation generates harmonics that fold back into the audible range as inharmonic noise. KERN WARM uses 2nd-order Baker ADAA combined with 2x oversampling for roughly 52 dB of alias suppression.

which saturation plugin has lower CPU usage?

both are efficient. KERN WARM targets under 3% CPU at 44.1 kHz with a single stereo instance, which includes its full spectral analysis pipeline. puffer:fish is similarly lightweight. neither will cause issues in a typical session.

references

a note from the developer

puffer:fish is a generous gift from Sonible. giving away a three-character saturation plugin for free raises the floor for everyone. that is good for producers and good for the industry.

the reason KERN WARM costs $29 is the spectral pipeline underneath the waveshaper. most saturation plugins just push your signal through a curve. KERN WARM looks at what frequencies are in your signal first, then decides how much saturation each frequency range should get. the Chebyshev math and Baker ADAA on top of that are there to make sure the harmonics it generates are musical, not digital artifacts.

this took three years of prototypes that sounded wrong before it sounded right. the hardest part was not the math. it was listening to the difference between “warm” and “muddy” on a hundred different sources and figuring out which spectral parameters drew the line between them.

if you have questions about any of this, or if you think i got something wrong, reach out at jonas@kernaudio.io.