French scientists developed a speech synthesizer that can turn mouth movements in words and logical sentences, bypassing the user’s voicebox #todaymagic
Scientists at INSERM and CNRS in Grenoble made this achievement possible by using nine sensors to capture movements of lips, tongue, jaw and soft palate. Afterwards, a neural network studied the data and turned it into a corresponding series of vocals and consonants which were fed to a vocoder. In return, this emitted the intended sentences in a robotic tone, but completely understandable by humans in sight.
As you can imagine, this could improve the lives of hundreds of paralysed people, allowing them to communicate more easily with their dear ones. For that to happen, there’s one condition, though: it has to be used as a building block for a brain-computer interface. This means scientists will have to work on a way to decode signals directly from the brain, without vocal tract movements.