Text-to-speech technology offers youngster a suitable voice
In May this year, Carte Blanche aired a programme on text-to-speech (TTS) technology and voice banking which featured an insert on Alexander Avenant, a learner at New Hope School in Pretoria. Alexander, at that stage, used the CSIR's Qfrency adult male Afrikaans TTS voice with augmentative and alternative communication software, because that was all that was available at that stage. In the interview with Carte Blanche, Alexander mentions that he does not like the fact that the voice sounds like an "Oom", and not like himself – a boy.
Three boys from Constantia Park Primary School saw the programme and decided to do something to help Alexander to have a voice that sounded more like a boy’s voice. They contacted New Hope School, and one of the speech–language therapists approached the CSIR HLT Research Group – which develops TTS voices for the South African market – to find out whether it would be possible for a child voice to be built using speech data that would be donated by the three boys.
Engineers from the CSIR HLT Research Group arranged for trial recordings to be done by the boys, and then selected the learner whose voice they determined had the best voice qualities for building a TTS voice. The learner then spent five days in a recording studio, reading preselected text which “covers the acoustic space” of the Afrikaans language. These recordings were then processed by the CSIR team, using its TTS system, to develop a child voice, which was put through its paces before being released.
The development of this TTS voice was dependent on having a large enough sample of Afrikaans text for the recordings. The text had to be carefully selected to be appropriate to the age of the learner doing the recordings; it also had to adequately cover the acoustic space of Afrikaans, and had to be open domain text, in order to be reused for commercial purposes. The South African Centre for Digital Language Resources (SADiLaR) plays an important role in ensuring that the corpora required to develop this and other speech technology, is widely available.
Seeing the footage of Alexander's first interaction with his new voice, was an exciting, proud, but also humbling moment for the CSIR team. It was wonderful to see the impact that this technology has made on Alexander's life. Congratulations to the boys from Constantia Park Primary School for their initiative and caring, and for the gift they have given Alexander. They are a true inspiration to all of us.
The new boy child voice, "Bennie", is available from the resellers of the CSIR's Qfrency TTS voices, Inclusive Solutions. More information is available at www.savoices.inclusivesolutions.co.za.