On the second day of the ongoing ABU RadioAsia conference, Anna Dixon, Senior Service Designer, ABC Innovations Lab gave a showcase presentation titled ‘The smartest things you’ve never seen’.
She discussed an exciting new project at the ABC Innovations Lab which started with the thought of being able to choose some articles from the ABC website and have them read to you as you drive into work or head out for a walk.
The pilot objectives were to develop a custom ABC synthetic voice to a level capable of reading out articles, evaluate voice capabilities by comparing a custom ABC branded voice and an off the shelf voice, apply synthetic voice in a text-to-speech trial to understand audience appetite and deliver insights into potential accessibility benefits as well as translation capabilities.
Dixon took the audience through the steps involved in developing this voice, giving a history of voice synthesis, which kicked off in the 70s with format synthesis, followed by a technique called concatenative synthesis, where recorded chunks of sound are reassembled in different ways and through to the modern voices of our smart home assistants and phones today produced by a few big tech players.
She discussed developing a voice using neural networks – a series of algorithms that endeavours to recognise underlying relationships in a set of data through a process that mimics the way the human brain operates.
She spoke about taking samples of three of their presenters and how they ended up developing an ABC virtual voice in partnership with Microsoft and applied it in a ‘text-to-speech’ trial at the ABC News website and app.
Discussing the viability of the pilot, Dixon said: “Is it practical or useless? We try to learn via tracking analytics, a survey that is attached to the feature, and some specialized usability testing with accessibility experts.”
This story first appeared on RadioInfo.asia