Most producers I’ve talked to open Serum 2, click through the five oscillator engines, land on Spectral, play one note, and move on. I get it. The Serum 2 spectral oscillator looks intimidating, and “spectral resynthesis” sounds like a DSP lecture.
But the idea underneath is simple. Once it clicks, you can turn almost any sample – a piano note, a spoken word, a struck coffee mug – into a playable instrument you can morph and reshape in ways a sampler never could.
We’ve shipped Serum preset packs at Monosounds since 2022, and since the 2.0 update I’ve built more patches on this engine than on anything except wavetable. Here’s what it actually does, in plain words, plus three recipes you can copy today.
Spectral Resynthesis in Plain Words
Every pitched sound is a stack of sine waves. A piano note isn’t one wave. It’s a fundamental plus dozens of overtones, each with its own volume, all changing over time.
Spectral resynthesis analyzes a sample and splits it into those sine components – Serum 2 calls them partials. Then it rebuilds the sound from the partials in real time instead of playing back the recording.
That’s the whole trick. A sampler hands you a frozen photo of a sound. Resynthesis hands you the recipe. And a recipe can be edited: remove ingredients, boost others, stretch the timing, change the order.
All of this happens in the frequency domain. You’re not editing audio anymore, you’re editing harmonics. Which is why this engine does things that are flat-out impossible with a normal sampler.
How the Serum 2 Spectral Oscillator Works
Switch any of the three oscillators to the Spectral engine and drag a sample onto it. Serum 2 analyzes the audio into spectral frames – snapshots of the partial content, moment by moment through the file.
From there, three abilities matter most.
Muting and Isolating Partials
You can edit the sound at the level of individual partials. Mute the bottom five and the fundamental disappears, leaving only the shimmer on top. Mute every even-numbered harmonic and you get a hollow, clarinet-like tone from any source.
My favorite move: solo a narrow band of upper partials from a piano. What’s left doesn’t sound like a piano at all. It sounds like glass.
Morphing Between Frames
The frame position control works like WT Pos does on a wavetable, except here the frames are the timeline of your sample. Scan through it slowly and a two-second vocal becomes a two-minute evolving pad. Freeze on one frame and you get an endless drone of that exact instant.
Assign an envelope or LFO to frame position and the sound moves on its own. This is where 80 percent of the magic lives.
The Transient Problem
Honest downside: resynthesis smooths out attacks. Sine waves are great at sustained tone and bad at clicks. So a resynthesized piano loses some hammer, and a resynthesized pluck can turn mushy.
Two fixes. Use a fast amp envelope to fake a sharper attack, or layer the raw sample on a second oscillator running the Sample engine, just for the first 50 milliseconds. I use the layering trick in almost every spectral patch I ship.
Where Spectral Fits Among the Five Engines
Serum 2 gives every oscillator five engines, and each one has a lane. I wrote a full breakdown in my guide to all five Serum 2 oscillator types, but here’s where the Serum 2 spectral oscillator sits in the lineup:
| Engine | Best for | Weak spot |
|---|---|---|
| Wavetable | Classic basses, leads, EDM staples | Static without heavy modulation |
| Sample | One-shots, exact playback, attack layers | It’s a fixed recording |
| Multisample (SFZ) | Realistic keys, strings, mapped kits | Bigger memory footprint |
| Granular | Gritty textures, stutters, motion | CPU climbs fast with grain count |
| Spectral | Morphing pads, harmonic surgery | Smears transients |
The closest cousin is granular, and people mix them up constantly. Granular chops the audio into tiny time-slices and replays them; spectral rebuilds the sound from sine partials. Granular keeps grit and texture, spectral keeps pitch and smoothness. I cover the grain side in my Serum 2 granular tutorial.
Three Sound Ideas You Can Build Today
Everything below uses the Serum 2 spectral oscillator, stock FX, and one sample you record or already own. Nothing exotic.
1. Airy Keys from a Piano Sample
Start with a single sustained piano note – C3, three or four seconds, no reverb on the recording.
- Osc A: Spectral engine, drag the piano note in.
- Mute the bottom five partials. The fundamental vanishes and the overtone shimmer stays.
- Assign Env 2 to frame position, attack around 800 ms, so the harmonics bloom after the note starts.
- Unison 3, detune 0.08.
- Osc B: Sample engine, same piano note, around -12 dB, just to bring back a touch of hammer.
- FX: hall reverb, size around 60, mix around 35.
The result is keys that sound like a piano remembered in a dream. We used variations of this on several key and pad patches in our free Serum 2 synth presets, so you can hear a finished version there.
2. Ghost Vocal Pad
Record one sung or spoken word, roughly one bar long at 120 BPM. Even a phone recording works – resynthesis hides a surprising amount of noise.
- Osc A: Spectral engine, drop the vocal in.
- Map LFO 1, triangle, synced to 8 bars, to frame position at around 35 percent depth. The word slowly dissolves and reforms.
- Mute the noisy partials at the very top to kill sibilance – the highest range does almost nothing for a pad anyway.
- Unison 5, detune 0.12, width full.
- Amp envelope: attack 1.2 s, release 2.5 s.
- FX: one of the new Serum 2 reverbs, big and dark, plus a light chorus.
You’ll recognize this trick in half the atmosphere patches in our free pack of 250+ Serum 2 presets. Open one, click the oscillator, and you can see exactly which partials we muted.
3. Metallic Pluck
Hit something metal – a mug, a radiator, a pan – and record one second of it. Metal rings with inharmonic partials, which is exactly what spectral analysis captures and a plain wavetable can’t.
- Osc A: Spectral engine, load the hit.
- Isolate a band of partials, roughly the 12th through the 30th. That cluster is where the clang lives.
- Env 1 on amp: attack 0, decay 400 ms, sustain 0.
- Send Env 1 to frame position too, small amount, so every note gets a little spectral sweep.
- FX: Bode frequency shifter at +7 Hz for extra metallic detune, then a 1/8-dotted ping-pong delay.
This kind of pluck sits beautifully in future bass and melodic house drops. If you want ready-made examples, the pluck and chord patches in Harmonic Horizons, our future bass Serum preset pack, lean on the same ideas.
Keeping the CPU Bill Reasonable
Spectral is heavier than wavetable. Rebuilding a sound from dozens of sine partials costs cycles, and stacking unison on top multiplies it.
What works for me: unison at 3-5 instead of 16, polyphony capped at 8 for pads, oversampling left at 2x unless I actually hear aliasing. And once a part is written, freeze or bounce the track. Xfer also publishes official CPU guidelines on the Serum 2 page if you want the full checklist.
One more trick: when a spectral patch is finished and no longer needs live morphing, use Resample to Oscillator to bake it into a wavetable. You keep the tone, drop the CPU cost, and can keep sculpting – I walk through that workflow in my post on making custom wavetables in Serum 2.
Serum 2 Spectral Oscillator FAQ
Is the spectral oscillator the same as granular?
No. Granular replays tiny time-slices of the audio, up to 256 grains at once, and keeps the raw texture. The Serum 2 spectral oscillator rebuilds the sound from sine partials, which gives you smoother results and per-harmonic editing. Same sample, very different outcomes.
Can I use any sample in the spectral oscillator?
Pretty much. Drag any WAV onto the oscillator and Serum 2 analyzes it. Clean, pitched, mostly-mono sources resynthesize best. Dense mixes and full drum loops turn to mush – that material belongs in the Sample or Granular engine instead.
Why does my spectral patch sound washed out?
Two usual causes. Transients get smoothed by resynthesis, so layer the raw sample underneath for attack. And piles of quiet upper partials add a hiss-like blur – mute the top range and the patch usually snaps into focus.
Does Serum 1 have a spectral oscillator?
No. It arrived with Serum 2 in March 2025. The upgrade is free for Serum 1 owners and your old presets carry over, so there’s no real reason to stay on version 1.
Steal Our Patches First
The fastest way to learn this engine isn’t reading. It’s reverse-engineering. If you want a shortcut, grab our free 250+ Serum 2 presets – no strings attached. Open the spectral patches, check which partials we muted and what’s driving frame position, and tear them apart until it makes sense.
I run Monosounds.studio and make the packs myself. Ten-plus years of producing, 97+ releases, 16,000+ downloads. Every Serum 2 preset we ship passes a CPU check on a mid-range laptop and gets used in a real track before release. If it does not survive a mix, it does not go in the pack.
More about Maxim & the studio →
