CereProc launches most advanced voice cloning technology in the market with CereVoice Me update

Innovative Scottish company brings together Deep Neural Networks and text-to-speech to help those suffering from a degenerative illness

Edinburgh, Scotland November 2018 – Text-to-speech experts CereProc today announced a breakthrough for its voice cloning system CereVoice Me, with an update powered by advanced Deep Neural Network (DNN) text-to-speech technology. The technology reduces the amount of data needed to clone a voice, improves speech output quality, and makes it possible to repair the speech of individuals suffering from a degenerative illness. This new technique was developed from experimental research conducted in the JFK Unsilenced project, creating a new JFK speech, as featured on The Times newspaper front page on the 16th of March 2018.

The JFK Unsilenced project blended two techniques of voice building - DNN and unit selection. Unit selection involves splicing together a real database of speech sounds, to capture the naturalness of speech. A DNN model was added to capture the unique speaking style of JFK. These experiments pushed CereProc’s technology forward and are now being applied to pure-DNN text-to-speech in all its technology including its voice cloning service. By using existing CereProc voices as a base, a text-to-speech version of an individual’s own voice can be created that reflects their identity and character.

Chris Pidcock, Chief Technical Officer, CereProc, commented: “Reducing the amount of recording time for individuals suffering from a degenerative illness is hugely beneficial because speaking for long durations can be exhausting. If some of the individual’s speech is degraded, we can now repair it using donor data from a similar existing voice.”

In addition to English, new languages have been added to the voice cloning system, including French, Spanish, Italian, Polish and Swedish, giving CereProc world-leading language coverage.

Paul Welham CEO, CereProc concluded: “CereProc is unique in offering Augmentative and Alternative Communication (AAC) companies existing voice technology free of charge for those with disabilities. We at CereProc are proud to advance the development of text-to-speech technology that can really touch the lives of individuals around the world. People who could become voiceless can instead keep their own voice.”

Ends…

Notes to the Editor:
• CereProc is the only text-to-speech company in the UK. We’re based in Edinburgh, Scotland, the home of advanced speech synthesis research, with a sales office in London.
• The CereProc team consists of leading speech experts who have extensive experience across the entire speech technology domain.
• CereProc was founded a decade ago to push the boundaries of speech technology, expand it into new markets, and use it to help people.
• In 2018, we used experimental deep learning technology on the JFK Unsilenced project to create the speech that JFK was due to give on the day he was assassinated in 1963.
• Also in 2018, CereProc gave radio journalist Jamie Dupree his voice back after it was taken by a rare medical disorder.
• In 2010, we used experimental technology for the first time to build a voice from past recordings for the film critic Roger Ebert.
• CereProc developed the world’s first Gaelic voice.

Contact: Sophie Henderson
Email: sophie@cereproc.com

CereProc Ltd
Codebase
Argyle House
3 Lady Lawson Street
Edinburgh
Midlothian
EH3 9DR
UK

Tel: +44 (0)131 516 8532
Website: www.cereproc.com
Twitter: @cereproc

This release was published on openPR.