Cross-synthesis imposes the spectral envelope of one sound onto another by multiplying their spectra

Cross-synthesis creates a hybrid sound by using the spectrum of one signal to shape another. In its most common form the magnitude spectrum of sound A scales, frequency-bin by frequency-bin, the magnitude spectrum of sound B — implemented by multiplying each point of spectrum A with the corresponding point of spectrum B. Because multiplication in the frequency domain equals convolution/filtering in the time domain, this is also called filtering by convolution: sound A acts as a time-varying filter imposed on sound B. It works best musically when one of the two sounds has a broad, noise-like bandwidth, so there is energy at every frequency for the other spectrum to sculpt — the mechanism behind classic vocoder effects (speech shaping a synth pad). A phase vocoder with two inputs performs this almost automatically. A variant borrows the magnitude of one sound and the phase of another to make a different hybrid.

Examples

A spoken-voice magnitude spectrum multiplied against a broadband synth chord makes the chord ‘talk’ — the vocoder effect. Convolving a drum loop with a vowel spectrum tints each hit with a formant colour.

Assessment

Explain why cross-synthesis by spectral multiplication produces its most striking results when one of the two sources is broadband (noise-like) rather than a pure tone, and name the classic effect this technique underlies.