Post-mix Vocoding and the making of All You Need Is Lunch

A six-minute audiovisual presentation, All You Need Is Lunch, was produced using a novel vocoding technique in which the spectrum of a sung word is altered from within a finished stereo mix, avoiding the need for blind source separation. In the piece, snippets of pop music tunes containing the word "love" are altered to say "lunch" instead, as in "where is lunch", "tainted lunch", "saving all my lunch for you", etc. To do this the utterance "lunch" is analyzed using an additive-synthesis model, and the musical recording to be altered is selectively filtered in specific, time-varying frequency ranges and left untouched elsewhere. The depth of alteration is frequency-dependent and time-varying. The target ("lunch") utterance must be time-morphed to fit optimally onto each individual source utterance ("love"). It proved particularly important, and often difficult, to either suppress or hide the sibilant portion of the "v" consonant. Since over 100 occurrences of the source word were altered, production tools were developed for editing and managing the many time-varying parameters that had to be chosen through critical listening and, ultimately, painstaking trial and error.