Invited workshop on synthesis of singing

Invited workshop on synthesis of singing

Johan Sundberg (KTH ROYAL INSTITUTE OF TECHNOLOGY, Sweden) Chair

ΤHURSDAY 18/19/2014, ODEON Starcity (12.10-16.20)

 

 

     Over the last decades much work has been spent in the field of synthesis of singing, a task that is both difficult and instructive.
This workshop will review some of the past and current achievements in this area.
     The difficulty with synthesizing singing is that as listeners we are inclined to interpret every detail of vocal sound in expressive terms. This is a habit probably emerging from our experiences of spoken communication. For example, if timbral or temporal accidents appear in synthesized singing, we are likely to detect it immediately and react negatively: "This doesn't sound natural".
     But working with synthesizing singing is also instructive, in that it offers a unique possibility to scientifically analyze the expressive code used in singing and also in music performance in general. The synthesis is a tool for analyzing what makes the sound of a singing voice sound like singing: "What is it that makes the singing sound natural rather than synthetic?" Furthermore, particular details of a synthesized sung performance can be varied systematically and evaluated by expert listeners, e.g., with respect to their effects on the expressivity. "What does the singer mean by doing this?"
     At the same time, synthesized singing can also stretch the limits of human expressivity. For example, human motion can be used to control the synthesis or the processing of vocal sound. Particularly if this synthesis or processing is realized in real time it can break the way to a new world of semi-vocal sound.
     There will be three presentations:

 

Marti Umbert

Expression Control in Singing Voice Synthesis:
Features, Approaches, Evaluation, and Challenges

 

     This presentation will give an overview of and demonstrate different attempts to model singing voice synthesis performances that have been made during the last decades.

     Martí Umbert, member of the Music Technology Group (MTG), lead by Dr. Xavier Serra, of the Department of Information and Communication Technologies at the Universitat Pompeu Fabra (UPF), Barcelona, is working in the Audio Signal Processing team lead by Dr. Jordi Bonada and Dr. Jordi Janer. After doing a 5 months research stay with the Vocaloid team at YAMAHA, in 2011 he started his PhD on expression control in singing voice
synthesis.

Carl Unander-Scharin and Ludvig Elblaus
Extending Operatic Voices in Live Performances:
Presentation of a range of artist-operated tools developed in collaboration with opera-singers for on-stage purposes

     Carl Unander-Scharin is composer of nine operas that have won international acclaim, is also a professional lyric tenor (e.g. at the Royal Swedish Opera in Stockholm), and has worked extensively with applying electronics in operatic practice. His research around Extending Opera, a joint project between KTH (Royal Institute of Technology) and The University College of Opera in Stockholm, has attracted international attention and has resulted in several research articles as well as novel embodied interactive instruments such as The Throat III, The Vocal Chorder, The Charged Room and The Throat for iPhone. Together with engineer/ artist Ludvig Elblaus he will talk about and demonstrate some of the systems they have developed that allow singers to use body and limb movements to modify and generate vocal sound.

 

Tod Machover

Searching for Singing's Soul: From Robot Operas to Vocal Vibrations

     Tod Machover is head of the MIT Media Lab's Opera of the Future group will talk about his work in the area of synthesized singing, which has negated traditional artistic and cultural boundaries. He has constructed the Hyperinstruments, a technology that uses smart computers to augment virtuosity. His opera Death and the Powers (2010) was developed by a creative team of international artists, designers, writers, and theatrical luminaries, as well as by an interdisciplinary team of Media Lab graduate and undergraduate students. It is scored for a small ensemble of specially designed Hyperinstruments, in which Powers features a robotic, animatronic stage—the first of its kind—that gradually "comes alive" as the opera's main character.