Kõneveeb ja Minu Hääl: uus kõnesünteesikeskkond ja -teenus
Kõneveeb and Minu Hääl: An interactive web environment for speech synthesis resources and a service for custom synthetic voice creation
Author(s): Meelis Mihkla, Indrek Hein, Indrek Kiissel, Jaan Pajupuu, Liisi Piits, Heete Sahkai, Hille Pajupuu, Rene Altrov, Elgar Kudritski, Liis Ermus, Egert Männisalu, Kristjan SulusteSubject(s): Sociolinguistics, Cognitive linguistics, Computational linguistics, Baltic Languages
Published by: Eesti Rakenduslingvistika Ühing (ERÜ)
Keywords: Speech; speech technology; speech corpora; machine learning; text voicing service; Estonian;
Summary/Abstract: Text-to-speech synthesis – a technology that converts written text into speech – has become part of everyday applications. This means that there is a general need for individuals to install speech synthesis in their devices and for companies to integrate it into their products and services. It is therefore important to make speech synthesis available for maximally easy uptake, especially for languages that are in danger of being dominated by English. The paper describes the interactive web environment Kõneveeb that is being developed to this end at the Institute of the Estonian Language. The purpose of the environment is to make the Institute’s Estonian speech synthesis resources and services easily available for individuals, companies, and developers alike. In addition to various free text voicing services, application interfaces, training corpora and training software, Kõneveeb offers a custom synthetic voice creation service, Minu Hääl. The service is intended for users who need a synthetic voice that is unique, represents a specific speaking style, or resembles their own voice. It allows the user to create a synthetic voice without any technical knowledge about speech synthesis. The service includes a dedicated program that enables an effortless recording of the donor voice and outputs a training corpus in the required format. After the user has uploaded the training corpus to Kõneveeb, three synthetic voices are automatically trained, using different synthesis methods. The resulting speech synthesisers can be used for text voicing either directly in Kõneveeb or installed in the user’s computer. Both the Kõneveeb environment and the Minu Hääl service will be constantly updated with new resources and possibilities.
Journal: Eesti Rakenduslingvistika Ühingu aastaraamat
- Issue Year: 2023
- Issue No: 19
- Page Range: 111-124
- Page Count: 14
- Language: Estonian