One of the most depressing things about the “land of tomorrow”-esque promises behind innovation–at least as the parent of a special needs child myself–is seeing how some of the most ordinary obstacles that any individual with a disability might face have yet to be overcome through tech. The clunky bionic suits of the ’80s that researchers put towards helping paraplegic people walk are barely any smaller–or more accessible–than they were back then. The technology behind the more recent “bionic” intuitive prosthetic arms still requires the gross national product of a small country to pay for it.


So when there is an innovation that is truly turning the tables on its predecessors, it should be shouted from the rooftops, as many news outlets are now doing with LipNet. This software, which has blown away previous lip reading software attempts with a mixture of Alphabet’s Deep Learning and autocorrect, has nearly 93% accuracy rate.

According to a paper published by the creators, “All existing [lip-reading approaches] perform only word classification, not sentence-level sequence prediction…. To the best of our knowledge, LipNet is the first lip-reading model to operate at sentence-level.”

That sentence level is what makes all the difference. If you’ve goofed around with early versions of speech-to-text software, you’ve probably had a good laugh at how some of your sentences turned out. That laugh isn’t nearly as funny when you’re trying to navigate daily life while relying on lip reading from the wide variety of people and dialogues around you.

Of course, support for the hearing impaired isn’t the primary goal for LipNet. It just happens to be a nice benefit that can support those with communication barriers. One of the major focuses is in better human-computer interface, meaning our computers are more likely to understand us when we speak without actually having to “hear” and process our words. There are, however, plenty of pundits who see the rights’ violations potential for this, as well as Big Brother-style overstepping by law enforcement and the government.