Pluck one string and you don't hear one note — you hear a whole ladder of them, the overtones, stacked in whole-number ratios. Their recipe is the instrument's voice. Drag the bars to set each overtone, tap PLAY to hear the timbre you built, and add sines until a pure tone turns into a buzzing saw.
This is additive synthesis on the real harmonic series: partial n has frequency n × the fundamental (2f, 3f, 4f…), and the drawn bars are their amplitudes. Tap PLAY and the page sums that many sine oscillators live through the Web Audio API — no samples, no recordings. The presets are the exact classical spectra: sawtooth = every harmonic at 1/n, square/clarinet = odd harmonics only at 1/n, pure = the fundamental alone. Fourier's theorem made audible: timbre is the amplitude of the overtones, and you can hear a pure tone become a buzz as you add them.
Real instrument tones also have inharmonic partials, noise, and time-varying spectra (the attack of a real string isn't a steady sum of sines); this is the idealised harmonic model with a fixed spectrum. The synthesis itself — n·f partials at the amplitudes you set — is exact and generated live, not a recording.