Behind Zero-G's English release stood years of university-backed computational research in Hamamatsu, Japan. In March 2000, Yamaha researcher Hideki Kenmochi kicked off an internal engineering project codenamed "Daisy." Kenmochi partnered with the Music Technology Group at Pompeu Fabra University in Barcelona, Spain, aiming to build a functional singing synthesizer using concatenative synthesis.
The technical challenge was formidable. Synthesizing spoken speech was already common, but singing required variable pitch control, emotional inflections, sustained vowels, and subtle micro-tonal shifts between consonants. Kenmochi's team chopped recorded phonetic fragments into tiny segments called dyphones and triphones. When a composer placed notes on a digital piano roll, the Vocaloid 1 engine stitched these tiny acoustic samples together, stretching and repitching them to match the target melody.
Yamaha had no interest in selling consumer voice packages directly. Kenmochi's team operated as an infrastructure developer, licensing the proprietary Vocaloid runtime engine to commercial sample companies. Zero-G took the first bet with Leon and Lola, following up six months later with Miriam, a software instrument voiced by South African singer Miriam Stockley. Crypton Future Media, a sound design vendor based in Sapporo, watched those early releases closely before deciding to test the same technology back home in Japan.
