Some of the best musical decisions ever made looked like mistakes at the time. Opera in a rock song. A hit single in 7/4. An album with almost no guitars from a guitar band.
I make sounds for a living, and I keep coming back to the same lesson: the tricks that feel wrong are often the ones people remember. So this post is a tour of famous rule-breaking moments that worked, plus concrete ways to steal the same thinking in your own sessions today.
No vague “be creative” advice. Real settings, real numbers, real workflows, including a few that use Serum 2’s newer engines, which honestly make this kind of experimentation easier than it’s ever been. If you want sounds to mess with while you read, grab the free Serum 2 presets pack first. It’s 250+ presets and it costs nothing.
Why breaking the rules works
Your listener’s brain is a prediction machine. Familiar chord progressions and song structures get processed on autopilot. The moment something breaks the pattern, attention snaps back to the music.
That’s the whole mechanism. A weird trick isn’t good because it’s weird. It’s good because it creates a moment of surprise inside something otherwise familiar. Too much chaos and people tune out. Too little and they scroll past.
The artists below understood that balance. Most of their “extravagant” decisions sit inside very normal songs.
Famous tricks that shouldn’t have worked
Strange instruments and sound sources
David Bowie’s Low (1977) is the classic case. He and Brian Eno pushed processed synths and drum treatments so hard that half the album barely sounds like a rock record. The harmonized snare sound on that album came from an Eventide harmonizer used in a way the manual never intended. It became one of the most copied drum sounds of the decade.
Radiohead did the same on Kid A (2000): a guitar band swapping guitars for Ondes Martenot, modular synths and cut-up vocal processing. Fans expected OK Computer part two. They got an electronic record instead, and it went to number one anyway.
You don’t need rare hardware for this. Record your kettle, a door hinge, a coin in a mug. Drop the recording into a sampler or, better, into a granular engine and stretch it 800 percent. I walk through exactly that workflow in my Serum 2 granular tutorial.
Structures that ignore the verse-chorus rulebook
“Bohemian Rhapsody” is six minutes long, has no repeated chorus, and stops dead for an opera section. Radio programmers said it was unplayable. It’s one of the most streamed songs of the 20th century.
The Beatles’ “Revolution 9” went further: eight minutes of tape loops and spoken fragments on the biggest pop album of 1968. Not a hit, but it dragged musique concrète into millions of homes and influenced generations of producers who sample and mangle audio today.
The practical version: take your finished track and delete the second chorus. Or start the song at the drop. Or put the bridge first. You’ll hate most of the results, but one arrangement in five will feel fresher than anything you’d have planned.
Harmony that leaves the key
“Yesterday” slips a chord from outside the key into bar two and nobody blinks, because the melody carries you across it. Nirvana’s “Smells Like Teen Spirit” runs F, B-flat, A-flat, D-flat, a progression that doesn’t sit cleanly in any single key, and it powered the biggest rock song of the 90s. John Coltrane built “Giant Steps” on key centers moving in major thirds, which broke jazz harmony so hard it became its own curriculum.
Steal it simply: write your loop in A minor, then move exactly one chord up or down a semitone. F major becomes E major or F-sharp major. Play the loop again. That one wrong chord is usually the hook.
Unexpected elements in the arrangement
Whitney Houston’s “I Will Always Love You” opens with 45 seconds of completely a cappella vocal. No label wanted a single with no instruments for the first minute. It spent 14 weeks at number one.
Pink Floyd’s “Money” grooves in 7/4, built around a tape loop of cash registers, and still got heavy radio play. Aphex Twin regularly replaces melody with texture and noise, and those tracks score films and sell out festivals.
In a modern session, the cheapest version of this trick is a foreign texture: a choir stab in a trap beat, a vinyl-noise loop in a clean pop mix, a stutter vocal where a synth lead should be. I keep a folder of chopped vocal textures from Platinum stutter vocal samples exactly for this: when a drop feels flat, one unexpected human sound usually fixes it.
Tricks you can run in your DAW tonight
Here’s my short list of experiments that pay off often enough to be worth the time. Each one takes under 15 minutes.
- The wrong-tempo bounce. Export your loop, re-import it, and force it to a tempo 15 to 20 BPM away without warping artifacts hidden. A 140 BPM trap melody at 122 BPM often turns into a lo-fi or R&B idea for free.
- Resample and destroy. Bounce your synth line to audio, pitch it down 12 semitones, add heavy saturation, then layer it quietly under the original. Instant weight, slightly wrong in a good way.
- Odd meter for one section only. Keep the verse in 4/4, write the pre-chorus in 7/8 by cutting one eighth note from the last bar. The chorus landing back in 4/4 hits twice as hard.
- Reverb on nothing. Send only the reverb tail of a sound into the mix, cut the dry signal completely. Ghost chords behind a dry beat create space without clutter.
- Random modulation roulette. Assign an LFO or random source to a parameter you’d never automate, like unison detune or wavetable position on a bass, at 5 to 10 percent depth. My Serum 2 modulation guide covers how to keep this musical instead of messy.
- Borrow from the wrong genre. Drop a phonk cowbell melody into a house track, or 808 glides into a pop ballad. Genre collisions are how half the new subgenres of the last decade started.
Serum 2 as an experiment machine
When I wrote the first version of this article, doing granular or spectral tricks meant buying separate plugins. Since the Serum 2 update in March 2025 (a free upgrade for Serum 1 owners), you get 3 oscillators and multiple synthesis engines in one window: wavetable, granular, spectral, sample and multisample.
That changes the economics of experimenting. Some quick recipes:
- Granulate your own voice. Record 5 seconds of yourself humming, load it into a granular oscillator, set grain size around 80 to 150 ms, density low, and slowly scan through the file. Free pad, completely yours.
- Spectral freeze for drones. The spectral engine can hold and blur a single moment of a sample into an endless texture. Great for intros and transitions; the spectral oscillator guide shows the exact settings.
- Draw a broken wavetable. Import any random WAV, even speech, as a wavetable and quantize the frames. Ugly on paper, often gold in a mix. See custom wavetables in Serum 2 for the import steps.
- Let the sequencer surprise you. Serum 2’s built-in sequencer with a random note pattern through a tight scale lock is the fastest idea generator I know.
If you’d rather start from finished sounds and reverse-engineer, the Colors preset pack is built around unusual bell and texture timbres, and Creative Plucks Vol.1 shows how far you can push a “normal” pluck before it becomes something else.
10 prompts for your next session
- Swap one core instrument for something from a different culture or century: gamelan, duduk, harpsichord, toy piano.
- Write the chorus melody first, then build a verse that deliberately avoids its notes.
- Replace your hi-hats with any non-drum sound: keys jangling, a lighter click, sliced breath noises.
- Pick one chord per song that doesn’t belong to the key. Just one.
- Automate tempo. Even a 2 BPM ramp into the chorus adds urgency nobody can name.
- Record one take on the worst mic you own and keep it if the vibe beats the clean take.
- Mute the kick for a full 8-bar section where everyone expects it.
- Build a drop from a single vocal chop and silence. Add elements back only if it stops working.
- Write in a mode you never use. Lydian for wonder, Phrygian for menace.
- Finish the ugly idea. The polished version of a strange idea beats the strange version of a polished idea.
None of these guarantee a hit. They guarantee you’ll stop making the same track twice, which is the actual problem most of us have.
FAQ: experimental music production tricks
How do I experiment without the track becoming unlistenable?
Change one variable at a time. Keep the structure normal if the sounds are strange, or keep the sounds normal if the structure is strange. Almost every famous example above breaks exactly one rule and plays the rest completely straight.
Do I need expensive gear to try these tricks?
No. A stock DAW, a phone microphone and one capable synth cover everything in this post. Serum 2 alone handles the granular, spectral and wavetable experiments, and if you’re new to it, start with my Serum 2 beginners guide before going down the rabbit hole.
What’s the fastest experimental trick that actually improves a track?
The one-semitone chord shift. Move a single chord in your progression up or down a half step. Ten seconds of work, and it instantly tells you whether your loop was boring because of the harmony or because of the sound design.
When should I stop experimenting and just finish the song?
Set a timer. I give myself 30 minutes of pure experimentation per session, then commit to whatever’s on the timeline. Endless options are worse for creativity than no options, and unfinished weird ideas help nobody.
If you want raw material to experiment on, start with the free Serum 2 presets. 250+ sounds, free download, and every trick in this post works on them. Break them, resample them, make them yours.
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 →
