This AI Switches Your Accent For Painless Communication Around The World
By Ell Ko, 20 Dec 2021
Image via Sanas
We all know of translation between languages. But what will happen when the language itself is shared between two or more people, but their accents complicate things?
As it turns out, the thing to do is to call on Sanas for help.
Created using a neural network and trained on recordings from professional voice actors, this artificial intelligence is capable of converting accents in spoken English to those you’d find in different parts of the world.
For example, a sentence can be “translated” from an American accent to a Filipino one, and Spanish accents to British, and vice versa.
“We want to develop this as a tool that helps people with human-to-human interaction, without hurting their cultural identities,” states Shawn Zhang, one of the founding trio. According to him, this can be used almost everywhere communication is needed: healthcare, entertainment, education, to name a few.
This runs locally on the user’s device, and can operate in real-time with 150 milliseconds of delay. The speech quality is described by IEEE Spectrum as similar to telephone audio, and it can be used alongside apps like Zoom, Skype, and WhatsApp.
It was developed by a trio from Stanford University, a project conceptualized at the beginning of the pandemic when school went online and students found themselves communicating largely through audio and video calls.
Without the privileges of face-to-face communication, it can be even more difficult to make yourself understood—something that the Chinese computer science major, a Russian AI-focused management science and engineering (MSE) major, and a Venezuelan business-focused MSE major understood too well.
“We knew from our own experience that forcing a different accent on yourself is uncomfortable… We thought if we could allow software to translate the accent [instead], we could let people speak naturally,” explains Andres Perez Soderi, the Venezuelan member of the trio.
Almost two years after they started working on the project, Sanas is now growing steadily. With a full-time engineering team of 14 people and around US$6 million in funding, it seems poised to change communication as we know it.
[via IEEE Spectrum, cover image via Sanas]