The Creative Speech Technology Network

People are increasingly expected in every-day like to interact with synthetic voices. These encounters are often not very satisfactory. The suggestion is that a major reason for this is that these voices do not sound 'right'. Synthetic voices have improved greatly in recent years in that they sound more human-like. Yet there remain two (conflicting) problems: On the one hand they are still not entirely human-like; most people spot the fact that they are listening to synthesizer within a few words.At the same time, even if it were possible to re-create a perfectly human-sounding voice, it would convey the wrong associations for what are computer-generated utterances. The computer does not have all of the other attributes of a person, such as intelligence, a sense of humour, empathy etc. Put another way, a computer voice should sound like a computer voice. The question is: how should that sound?The objective of this network is to bring together people who know about (human) voices and how to use and manipulate them - from an artistic background - with technicians who know how to generate synthetic voices. By establishing and developing a language through which these disparate people can communicate, it will be possible to create new voices for specific purposes. The focus of the work will be on creating artistically inspired artefacts which will challenge the potentials of synthetic voice production. Those artefacts will also be used to disseminate the results of the work - to expose it to members of the public. This will share the results of the research as well as allowing the final evaluation of the voices to be given to the people who matter - those who have to put up with this kind of interaction on a daily basis.

Anyone who interacts with any systems using a synthetic voice should benefit from this work. That could be almost anyone in the current environment, but specific applications would include: call centres, interactive (particularly phone-based) booking and shopping systems, some public announcements (e.g. airports and railway stations), voice-based guides, as used in galleries and museums. Through the commitment to public dissemination, these every-day users will become more aware of the possibilities and to expect better use of speech in their encounters with these systems. Particular benefit should accrue to those who use synthetic speech as a means of expressive communication, usually as a result of their disability. There will also be a benefit to those who rely on synthetic speech as a result of visual impairments. This will be realized mainly through members the network, both academic and industrial. Artists will benefit not only from new possibilities in voice synthesis but also from an enhanced understanding of voice and how it works. New artistic pieces will be created - and exhibited - within the network, but there would also be expected to be new opportunities beyond its existence. Ultimately speech synthesizer manufacturers should benefit from the ability to create better products. Those who research speech synthesis will benefit from new insights which are based on the notion that human verisimilitude is not the only worthy quality. Designers and developers of human-machine interactions will benefit from new possibilities for communication. By the very nature of a network, there will be dissemination within the broad community of interested parties, some of whom have already been identified, others who will be expected to join in with the network when it has been created. The network will be focused on the creation of artefacts which will be the basis of dissemination particularly to the general public, as well as the research community. Other 'traditional' forms of research dissemination - papers, conferences and websites will also be exploited to share the benefits with the research communities beyond the network.