ReadSpeaker has put its expertise to work to develop state-of-the-art, clear and slow speech synthesis for the Swedish Aphasia Association.

ReadSpeaker’s Deep Neural Network Technology in action for Afasiförbundet

ReadSpeaker produced a pilot project in 2018-2019 in collaboration with the Afasiförbundet, with the PTS, the Swedish Post and Telecom Agency supporting the project. Afasiförbundet members with aphasia and language impairment, as well as people with other cognitive disabilities, participated in testing the slow speech synthesis voice. After working through the input and feedback gathered, and using Artificial Intelligence (AI) in development, the pilot voice became a fully-fledged product of slow speech synthesis in the spring of 2020. The new Swedish voice was developed with ReadSpeaker’s high quality Deep Neural Network (DNN) voice technology, using only 3 hours of speech, and is a compelling example of ReadSpeaker’s DNN voice technology in action. 

Voice is an important support tool for people who have difficulty reading traditional text, such as people with aphasia, visual impairment, acquired brain injury, ADHD, autism, dyslexia and language disorders. The project received support and involvement from the Post and Telecom Agency, the Aphasia Association, the Dyslexia Association, the National Association for Attention Disorders, the National Association for the Visually Impaired National, and the Brain Injury Association, the Personal Injury Association and the Autism and Asperger’s Association.

Speech synthesis and Artificial Intelligence

Speech synthesis is traditionally created with pre-recorded speech, which is then cut into small parts and put back together in different ways to form all possible words and sentences. Individuals then are given some limited flexibility to choose the speed of the voice, to help with comprehension. 

With the help of AI technology, ReadSpeaker has built the first slow synthetic voice where the voice is recorded from the beginning at a slower pace, which gives a clearer and more concise voice, with greater flexibility to meet a listener’s needs.  

Equal access for all a real possibility

What is more, with ReadSpeaker’s innovative deep neural network technology, only 3 hours of recordings was needed to create the new voice. This revolution in voice creation allows for quicker turn around, and more affordable voices, which opens voice technology up to many more organizations and their audiences, making it possible to provide equal accessible access for all.  

ReadSpeaker’s slow voice is a revolution and is setting the new standard for voice synthesis.

Contact us for more information about Deep Neural Network voice technology.

Contact Us