February 2, 2006.
Music 209 : Calendar : Week 3Guest lecture on concatenative speech synthesis. Recording vocals in a recording studio. Concatenative speech synthesis.
The lecture began with an excellent review of speech related technologies by guest lecturer Robert Ecklund.
We then began a discussion of modern techniques for recording lead vocals in pop songs. Much of the material was adapted from this review article in Sound on Sound and this review article in Mix Magazine.
We discussed pitch correction technology: see this article for a discussion of using Auto-Tune in a studio environment, and see this web page for the Auto-Tune audio examples presented in class.
We discussed vocal modeling as an alternative to traditional equalization -- see this review of a voice modeling product for a good discussion of how these algorithms work.
We ended the lecture segment on recording vocals with a discussion about the quality of Alanis Morissette's intonation on her album Jagged Little Pill, and how this issue did not prevent the album from selling 30 million copies. To read about the recording of this album, see this article.
We discussed the Vocaloid singing synthesis system in depth. These review articles provide a good introduction to the strengths and weaknesses of their system: [article1] [article2] [article3] [article4] [article5]. The company-produced demos of the Leon, Lola, and Miriam voices can be heard via MP3s linked on this webpage. This webpage has user-contributed Vocaloid demos.
A paper on an interesting research system for singing synthesis system that addresses vibrato in a novel way is available here. A good paper on diphone synthesis methods is available here.
We also discussed products for phrase-level synthesis of children choirs. See this product review and listen to demos here.
We also discussed several project ideas that involved vocals. See this web page for details.