Zero G announces Vocaloid products!

[elektro80]

Launched at Frankfurt Musikmesse 2003, Yamaha’s Vocaloid ‘singing synthesis’ technology created a fair amount of excitement, promising as it did to usher in the age of the virtual vocalist. Now Zero G have announced the imminent release of the first software products built on Yamaha’s technology — Vocaloid Leon and Vocaloid Lola, male and female virtual vocalists for the PC.

Those who shudder at the thought of such things and mutter darkly about the decline of ‘real’ music won’t be comforted by Zero G’s boast that “the last bastion of human musical expression — the singing voice — has been realistically harnessed in synthesis”. But, in truth, these products are aimed at the same people who bought Steinberg’s Virtual Guitarist — home users who want to add a little guitar (or vocals) to their compositions but lack the ability to play it themselves. And, like Virtual Guitarist, the Vocaloid libraries are, at root, samples of a real musician.

The biggest obstacle to voice synthesis in the past has been finding a way to realistically articulate individual words. Vocaloid’s strategy is to sample a real singer in minute detail, recording all possible phonemes and transitions between every different combination of syllables and sounds. Expressive artefacts, like vibrato, swell, pitch-bend and glide, are sampled separately and rendered as freely assignable expression templates. The resulting data can then be processed by the Vocaloid synthesis engine (Yamaha are naturally not forthcoming with the details of how this works!) to produce a virtual rendition of the melody and lyrics of the user’s choosing.

Naturally, the process of rendering the synthesized vocal is very CPU-intensive, and Zero G recommend a Pentium 4 PC running Windows XP, with 1GB RAM or more. Because of the level of detail required by Vocaloid libraries, each one is necessarily limited to the sex, style and language of the singer sampled. Leon and Lola are male and female libraries equipped with the full range of English phonemes and designed for soul and dance styles. Lyrics and notes can be entered via a piano-roll editor which can handle up to 16 vocal tracks, and expression templates are applied via a drag-and-drop interface. The software will also allow the user to import and export Standard MIDI Files.

Leon and Lola, which will be launched at the California Winter NAMM show in January, are sold separately and cost £199.95. A third library, sampled from the voice of Miriam Stockley (Vocaloid Miriam), will be launched at the Frankfurt Musikmesse in March 2004.

http://www.timespace.com

[Cyxeris]

I wonder how difficult it would be, in this age, to synthesize the mechanics behind a real human voice. You have oscillators (vocal cords), you have filters and eq (flesh, mouth, tongue, lips), you have breath. Certainly someone somewhere has got to be on the heels of such a technology, perhaps Bell Labs or Caltech or the mafia or someone.

Cyx

[seraph]

[seraph]

Read also what Jeff Harrington had to say about his piece "AgnusDeiWave"