Audio Synthesis Recipes
A field guide to building instruments from oscillators: the smallest set of ingredients that carries each instrument's recognizable identity, organized by instrument class.
What gives the sound produced by a piano it’s ‘pianitude’, distinct from the ‘oarganicity’ of an organ, the ‘violinness’ of a violin, or the ‘harpacy’ of a harp? This page explores the ingredients that give each instrument and instrument class it’s acoustic identity, through minimal identifiable synthesis recipes.
Every entry is a recipe, not a reproduction: the smallest set of ingredients that carries an instrument’s recognizable identity. The coloured pill on each recipe names its engine; the strip under it sketches the recipe symbolically — amplitude envelope, harmonic spectrum, and spectral centroid (brightness) over time relative to the note’s pitch. Items marked Lab are playable in the companion Instrument Anatomy tool (Coming Soon).
How to read a recipe
Seven engines cover almost everything. The pill colour on each instrument tells you which one; picking it is most of the battle, and the rest is choosing partials, envelope shape, and one or two modulations.
- Subtractive subtractive — Start with a harmonically rich wave (saw/square/pulse) and carve it with filters. The workhorse for anything driven and harmonic: strings, brass, reeds.
- Additive additive — Sum sine partials directly, each with its own level and envelope. Exact spectral control — organs, bells, tuned bars, anything where you want named partials.
- Frequency Mod fm — One oscillator modulates another’s pitch. Cheap inharmonic and metallic spectra — electric pianos, bells, clangs, bite on an attack.
- Karplus–Strong karplus–strong — A noise burst trapped in a tuned feedback delay = a real plucked/struck string, with the natural bright→dark decay built in.
- Source–Filter source–filter — A buzzy source through fixed resonant peaks (formants). The voice, and the “body” of any instrument whose resonances don’t move with pitch.
- Noise + Envelope noise+env — Filtered noise shaped by an amplitude (and pitch) envelope. Cymbals, shakers, snares, breath, bow scrape.
- Inharmonic additive / Ring inharmonic — Sines or ring-modulated pairs placed at non-integer ratios. Bells, gongs, glass, anything that chimes rather than sings.
The three questions that pick an engine
- Driven or decaying? Continuously fed energy (bowed, blown, electric organ) gives a flat sustain. Struck or plucked gives no sustain — it rings out. This is the largest single split.
- Harmonic or inharmonic? Strings & air columns lock to integer partials. Stiff bars, plates and bells do not — their partials are stretched, which kills clear pitch and reads as “metallic” or “woody.”
- Pitched or noise? Membranes and metals with dense inharmonic modes are effectively filtered noise with an envelope; only a few (timpani, steelpan) are tuned enough to play melodies.
Reading the strip: the envelope trace is loudness over the note’s life; the spectrum bars are which partials are present (gaps, inharmonic spacing, or a noise band all show here); the centroid curve is brightness over time against a dashed line marking the note’s pitch — falling means bright→dark, rising means a brightness bloom. Plots are symbolic, sized to show relative differences, not measured.
Struck & Plucked Strings
Chordophone · decaysEnergy is deposited once and the string rings out — no sustain, a bright→dark decay (high partials die first), and slight inharmonicity from string stiffness. The onset (hammer thunk vs. pick scrape vs. quill) and the decay length separate them.
Piano
Percussive onset + ring-out + faintly stretched partials. That inharmonic stretch is the “warmth” that isn’t in a pure-harmonic tone.
9+ stretched partials (B≈0.001), ~4 ms strike + broadband thunk, per-partial decay, double-decay tail; velocity→loudness & brightness together.
Electric piano (Rhodes/Wurli)
A glassy inharmonic bell-bark at onset melting into a near-sine body. Wurlitzer = add a touch of reedy second-harmonic grit.
Carrier sine + modulator at ~14×, index decays in ~150 ms (the bark), body sine decays over 2–3 s; velocity→index.
Harpsichord
A sharp quill click over a thin, bright, dynamically-flat pluck — the absence of velocity response is half its character.
Bright plucked string (KS or rich saw + fast attack), prominent pick/quill click, fast-ish decay, lid/case formants; velocity does nothing (the quill plucks the same every time).
Clavinet
Funky, percussive, almost nasal twang — the pickup narrows the spectrum and the attack is everything.
Plucked/struck string through a magnetic pickup: narrow bright spectrum, very fast attack, short decay, strong upper-mid; often distorted/wah’d.
Acoustic guitar (steel/nylon)
Pick attack + pluck-position comb filtering + wooden body resonance, dulling as it rings.
KS: 12 ms noise burst → tuned delay + lossy lowpass loop; pluck position notches certain harmonics (comb); body + soundhole formants (~100–200 Hz). Nylon = softer burst, darker loop.
Electric guitar (clean / dist.)
Clean: bell-like sustain. Distorted: harmonic saturation + near-infinite sustain — the clipping, not the string, is the timbre.
Plucked string, narrower pickup spectrum, long sustain (solid body loses little energy); distortion = waveshaping/clipping that adds harmonics and compresses dynamics; cabinet formants.
Bass guitar
Fat fundamental with a fingered thump or slap snap on the attack.
Low plucked string, strong fundamental + 2nd, short bright attack click, medium decay; finger vs. pick vs. slap changes the onset noise.
Harp
A soft pluck with a long, pure, harmonic ring — cleaner and less percussive than a piano.
Like piano but gentler onset (no hammer), very low inharmonicity (flexible strings), long clean decay, no sustain pedal damping by default.
Mandolin / 12-string
Bright pluck with shimmer from paired strings beating against each other.
Paired strings detuned a few cents → chorus/beating; bright fast pluck; tremolo picking = rapid re-triggers.
Banjo
Twangy, percussive, fast-dying — the drumhead resonance is the giveaway.
Plucked string coupled to a drum head: very bright, very fast decay, strong attack transient, slight membrane “plonk”.
Sitar
A buzzing, sustaining shimmer — the sympathetic strings keep ringing under the melody.
Main plucked string + a flat “jawari” bridge that buzzes (waveshaping/added high harmonics) + many sympathetic strings (extra detuned KS voices ringing).
Pizzicato strings
A dry, woody pluck that dies fast — the orchestral string body without the bow.
Short bowed-string body but plucked: fast attack, quick decay, body formants, slight pluck noise.
Bowed Strings
Chordophone · drivenThe bow feeds energy continuously, so the tone sustains and can swell. Stick-slip friction gives a near-sawtooth spectrum, and mode-locking forces the partials onto exact integers (no inharmonicity). Vibrato and, for sections, ensemble detuning dominate the character.
Solo violin
Singing sustained saw with expressive vibrato; bow position (sul tasto→dark, sul ponticello→glassy) = move the filter.
Sawtooth → gentle lowpass + body formants (~300–3 kHz); ~80–120 ms bow attack; vibrato ~5–6 Hz, delayed; bow pressure→brightness.
Viola / cello / double bass
The bowed-saw identity shifted down with a rounder, chestier body resonance.
Same engine, lower range, larger body formants moved down, slightly slower attack, woodier lowpass.
String section / ensemble
The shimmer of many players beating — detune + numbers, not a single richer voice.
4–9 detuned saws (±3–12 cents) + timing/phase spread; slow swell; collective vibrato.
Tremolo / col legno / harmonics
Articulation tricks: flutter, wood-clatter, or a glassy whistle over the same bowed core.
Tremolo = fast amplitude/retrigger flutter; col legno = mostly bow-wood noise (N+E); natural harmonics = near-sine at a node.
Hurdy-gurdy
A droning bowed sound with a percussive buzz ticking underneath.
Bowed (wheel) drone saws + a buzzing bridge (rhythmic waveshaping) + sympathetic drones.
Brass
Aerophone · lip-reed · drivenLips buzz into a tube; the tone brightens as you blow harder. The signature is a filter that opens just behind the volume, with bloom speed and height tracking effort. Mutes add strong fixed formants.
Trumpet
The brightness bloom coupled to loudness is the whole identity — play soft = dull, hard = blazing.
Sawtooth; fast amp attack; filter env attack slightly LONGER than amp → brightness bloom; velocity→cutoff; delayed vibrato; optional FM/noise rasp burst at onset.
Trombone
Brass bloom plus glissando slides connecting pitches.
As trumpet, lower, with portamento (slide) between notes; broader bore = slightly rounder.
French horn
Rounder, more distant brass — the bloom is softer and the top is rolled off.
Saw with a darker lowpass (mellower), slower bloom, hand-in-bell = a gentle high-cut formant; wide warm vibrato.
Tuba / euphonium
Brass identity at the bottom: round, slow, weighty.
Low saw, fat fundamental, slow gentle bloom, broad body.
Muted (straight / Harmon / cup)
A pinched, vocal “wah” — the mute imposes a fixed resonance the open horn doesn’t have.
Brass core + a strong fixed bandpass formant; Harmon = a sharp peak you can sweep (wah); nasal, thin, buzzy.
Woodwinds
Aerophone · drivenAir columns, but the bore shape sets the harmonics: a closed cylinder (clarinet) keeps only odd harmonics; a cone (sax, oboe, flute-ish) keeps the full series. Nearly all carry audible breath noise, strongest at onset.
Flute
Almost a pure tone — the breath noise is what makes it a flute and not a sine.
Triangle/sine (few harmonics) + filtered breath noise loudest at attack; soft attack; vibrato ~5 Hz.
Piccolo / recorder / pan flute
Variations on near-sine + air; the amount and color of breath separates them.
Flute engine: piccolo = octave up, more breath; recorder = purer, less breath; pan flute = strong breathy chiff at onset.
Ocarina
The purest wind voice — close to a flutey sine with almost no overtones.
Vessel (Helmholtz) resonator → essentially a single sine + tiny 2nd harmonic, gentle attack, light breath.
Shakuhachi
A breathy, vocal, bending flute — the noise and pitch inflection carry it.
Flute core with heavy breath, pitch bends (meri/kari), and noisy attacks.
Clarinet
Hollow and woody because even harmonics are missing — the square wave is the shortcut.
Square wave (= odd harmonics), gentle lowpass, medium reedy attack, slight breath, subtle vibrato.
Saxophone
Reedy and vocal — the formants and breath make it sing; full harmonic series, unlike the clarinet.
Sawtooth (conical → full series) + strong vocal formants + plenty of breath; medium attack; growl = add noise/FM.
Oboe / English horn
Pinched and nasal — a tight formant on a reedy double-reed buzz.
Narrow pulse or saw + a strong nasal formant pair (~1–3 kHz), thin and bright, quick reedy attack, vibrato.
Bassoon
The dark, woody, slightly comic double-reed — oboe character moved low.
Low narrow-pulse/saw + low formants, woody, soft attack.
Harmonica / accordion / harmonium
Organ-like flat sustain with a reedy buzz and, for accordion, that detuned musette beating.
Free reeds: saw/square stack, near-instant on, flat sustain, fast off; accordion = two reeds detuned (musette tremolo); vibrato via bellows.
Organs & Pipes
Aerophone/electromech · drivenContinuously driven and dead-flat: instant on, steady hold, fast off, with little or no envelope evolution. Tone comes from which partials (stops/drawbars) are mixed, not from dynamics.
Hammond / tonewheel
Literal additive synthesis — drawbars are harmonic faders; the absence of dynamics is the identity.
Sum of near-sine partials at drawbar ratios (16′,8′,5⅓′,4′…); rectangular envelope; key-click transient; scanner vibrato / Leslie chorus.
Pipe organ (flue stops)
Massive, static, additive — voiced by stop selection, with a breathy chiff as pipes speak.
Stacked stops at octave/fifth ratios; flue = flutey near-sine + a little chiff at onset; principals = brighter saw-ish.
Pipe organ (reed stops)
The snarling, buzzy organ color — a reed, not a flue.
Buzzy saw/pulse + formant (the resonator), flat sustain.
Combo organ (Vox / Farfisa)
Cheap, bright, reedy 60s organ — complex-tone partials instead of clean sines.
Drawbars of complex (not pure) tones → brighter, buzzier, thinner; fast vibrato.
Voice
Source–filter · drivenThe textbook source–filter instrument: vocal folds make a buzzy glottal pulse (rich in harmonics), and the vocal tract imposes formants (F1, F2, F3) that select the vowel. Move the formants, not the pitch, to change vowel.
Sung vowel
Pitch and vowel are independent: harmonics set pitch, formants set the vowel.
Sawtooth/pulse glottal source → 3 high-Q bandpass formants. “ah” ≈ F1 700 / F2 1200; “ee” ≈ 300 / 2300; “oo” ≈ 300 / 870. Vibrato ~5–6 Hz.
Choir / vocal pad
Vowel formants plus ensemble shimmer — a section of singers, not one.
Several detuned vowel-voices (ensemble) + slow attack + collective vibrato + reverb.
Whisper / breath
Formants without pitch → unpitched, breathy vowels.
Filtered noise (no glottal source) shaped by the same formants.
Overtone / throat singing
A whistling harmonic floating over a drone — one formant picking out partials.
Drone source + one extremely narrow sweepable formant that isolates single upper harmonics.
Tuned Percussion — Bars & Plates
Idiophone · struck · decaysStruck stiff bars. A free bar’s natural overtones are inharmonic, so makers undercut the bar to pull one overtone into tune and add tuned resonators. Wood = short decay; metal = long. The overtone ratio is the family fingerprint.
Marimba
Dark, round, mellow — the 1:4 overtone sits with the orchestra, so it blends and dies quickly.
Fundamental + overtone at ≈4× (two octaves, “octave tuning”) + weak ~9.2×; soft mallet; tube resonator on the fundamental; fast wood decay.
Xylophone
Bright and cutting — the 1:3 overtone (clarinet-like) makes it ring out over everything.
Fundamental + overtone at ≈3× (a twelfth, “quint tuning”) + a spike near 5×/7×; harder mallet; short decay.
Vibraphone
Shimmering metallic sustain with that rotating-disc tremolo — the one tuned-bar instrument that breathes.
Aluminium bars, overtone at ≈4×, long decay; motor-driven disc valves → amplitude tremolo; pedal sustain.
Glockenspiel
Tiny bright bell-like ping — high enough that the inharmonicity reads as sparkle.
Small steel bars: bright, high, partials not well-tuned (audibly inharmonic), short metallic ping.
Celesta
A music-box-like bell tone with a soft, rounded onset.
Hammered steel plates over resonators: glockenspiel-ish but soft, sweet, gentle attack, longer decay.
Kalimba / mbira
Round metallic pluck with a faint inharmonic buzz — thumb-piano warmth.
Plucked metal tine: strong fundamental + a couple of slightly inharmonic partials + a “buzz” overtone; medium decay, body resonance.
Music box
A small, bright, slightly inharmonic chime — sparse and toy-like.
Struck steel comb tine: bright near-bell partials, fast attack, medium decay, thin and delicate.
Steelpan
Unusual for metal: a genuinely harmonic, singing, shimmering struck tone.
Hammered pan note tuned so the first few harmonics are real (1,2,3 strong) + metallic shimmer; medium decay.
Glass harmonica
An ethereal, almost pure-sine sustained tone with a glassy edge.
Rubbed glass: near-sine fundamental + faint partials, slow swell (driven, not struck), pure.
Bells & Gongs
Idiophone · struck · inharmonicStrongly inharmonic metal. A church bell is deliberately tuned to five partials — hum, prime, tierce, quint, nominal — but the tierce is a minor third above the strike note, which is why bells sound plaintive. Gongs/cymbals abandon tuning entirely: dense inharmonic clouds = pitched noise.
Church / tower bell
The minor-third tierce over the strike note is the melancholy “bell” color; the hum rings on after.
Additive: hum (×0.5), prime/strike (×1), tierce (×1.2 = minor 3rd), quint (×1.5), nominal (×2) + higher inharmonics; sharp strike, very long uneven decay (hum lingers).
Tubular bells / chimes
A clear chimed pitch that isn’t physically the lowest partial — your ear infers it from the upper modes.
Upper modes tuned near 2:3:4 so a strike note emerges (no real fundamental there); bright metallic attack, long ring.
Handbell / crotales
Same inharmonic bell DNA, scaled in size and ring time.
Bell partial set, shorter (handbell) or very high & sustained (crotales).
Cowbell
A clanky, pitch-ambiguous hit — a few dissonant metal modes.
Two or three clashing inharmonic partials, fast decay, no clear pitch.
Triangle
A bright sustained metallic ting with no definite pitch.
Very dense high inharmonic partials ≈ bright metallic noise + faint pitch, long shimmer.
Gong / tam-tam
A swelling, shimmering metallic cloud — the bloom after the hit is the signature.
Dozens of inharmonic partials + noise + a low mallet thump; soft mallet → the sound builds and shimmers after the strike (energy spreads between modes); huge decay.
Cymbal (crash/ride)
Broadband metallic noise; ring-mod gives the inharmonic “clang” a filter alone can’t.
Bandpassed white noise + a few ring-modulated square pairs for metallic edge; fast attack, long wash (crash) or pingy stick attack + sustain (ride).
Hi-hat (closed/open)
A short, bright, metallic chick — cymbal recipe with a fast gate.
Highpassed noise + ring-mod metal; closed = ~40 ms gate, open = long; tight transient.
Drums
Membranophone · struckA struck membrane’s modes are inharmonic and mostly unpitched; pitch comes from a fast downward pitch envelope plus body resonance. Only timpani (kettle + air loading) tunes its modes near-harmonic enough to play melodies.
Timpani
The one tunable drum — near-harmonic loaded modes give it a real, if dark, pitch.
Kettle loads the membrane so prominent modes approach ~1:1.5:2:2.4 (definite pitch); sine-ish partials + short pitch drop + mallet noise; long resonant decay.
Kick / bass drum
The pitch-drop “thump” — a falling sine is 90% of a kick.
Sine/triangle with a fast downward pitch envelope (e.g., 120→40 Hz in ~50 ms) + a click transient; short body decay.
Tom-tom
A tuned “doom” with an audible downward bend.
Like kick but higher and more resonant: pitched sine + downward bend + a little head noise + body ring.
Snare
Tone + noise rattle together; the noise band is the snare wires.
A tonal shell (two detuned sines, fast pitch drop) + a burst of bright noise (the snares) gated to ~150 ms.
Congas / bongos / tabla
Hand-drum tone — pitched membrane with slap articulation; tabla is the most melodic.
Sharp tonal slap (pitched, fast decay) + open/muted variants; tabla adds a pronounced pitch bend and clear tuning.
Frame drum / taiko
A deep, dry boom — minimal pitch, maximal body.
Low broadband boom: noise + low sine, fast decay, big body.
Unpitched Percussion
Noise + envelopeNo pitch to speak of — these are noise (or a click) through a filter and a tight envelope. Identity lives in the noise color and the envelope shape (sharp tick vs. swish vs. rattle).
Woodblock / claves
A dry tonal “tock” — one short resonant pop.
A short resonant click: bandpassed noise or a quick sine ping, ~30–60 ms, high-mid.
Shaker / maracas / cabasa
A filtered noise “chh” — color and envelope are everything.
Highpassed noise burst, ~30–80 ms, soft attack; rhythmic re-triggering makes the groove.
Tambourine
A jingly metallic swish over a soft hit.
Bright metallic noise (many tiny jingles) + a head-hit transient; longer shimmer on shakes.
Clap / snap
Noise with a stuttered multi-tap attack — that’s what says “clap” not “snare”.
Several closely-spaced noise bursts (the “many hands” flam) into a short reverb tail.
Guiro / ratchet / cabasa scrape
A rasping, ratcheting texture — pulsed noise.
Rhythmically amplitude-modulated noise (a train of grains).
References
- Grey (1977) & McAdams — timbre space, attack transients, spectral centroid — mcgill.ca/mpcl
- Gordon Reid, Synth Secrets (Sound on Sound) — subtractive synthesis of real instruments — soundonsound.com
- Synth Secrets — synthesizing brass (filter bloom) — soundonsound.com
- Bill Hibbert, The Sound of Bells — bell partials, hum/prime/tierce tuning — hibberts.co.uk
- Strike tone & the perceived pitch of bells — hibberts.co.uk/strike.htm
- Stanford CCRMA — percussion: bar & plate overtone tuning — ccrma.stanford.edu
- Euphonics — marimbas & xylophones (measured bar ratios) — euphonics.org
- Yamaha — marimba vs. xylophone tuning (octave vs. quint) — hub.yamaha.com
- UNSW Music Acoustics — saxophone (conical bore, breath floor) — newt.phys.unsw.edu.au
- Wikipedia — Inharmonicity (stretched piano partials, “warmth”) — en.wikipedia.org
- Wikipedia — Hammond organ (additive drawbars) — en.wikipedia.org
- Wikipedia — Formant (vowel F1/F2 source–filter) — en.wikipedia.org
Recipes are starting points — real instruments are messier, and the fun is in the deviations.
Disclaimer: This field guide was researched and compiled with AI assistance. The synthesis recipes are simplified starting points, not measured reproductions.