Lipreading is the task of decoding text from the movement of a speaker's mouth. Traditional approaches separated the problem into two stages: designing or learning visual features, and prediction. More recent deep lipreading approaches are end-to-end trainable (Wand et al., 2016; Chung & Zisserman, 2016a). The Department of Computer Science at Oxford University has developed an app that uses AI software to read lips more accurately that human beings. The software is called LipNet and has a 92% accuracy rating. Indeed it scores 13% higher than the average lip-reading software. It does this by focussing on sentence construction, rather than deciphering one word at a time. The programme is also smart, and learns to become more accurate over time. The technology has exciting potential applications both for helping deaf individuals communicate and improving smartphone voice commands.