TalTech doctoral candidate creates web solution that clones your voice deceptively accurately

TalTech doctoral candidate Aivo Olev has created Jutusta.ee, an Estonian-language speech technology web solution that converts speech recordings to text, reads written text aloud in a human voice, and clones the user's own voice. The result is so convincing that even close relatives cannot tell the original from the copy.

TalTech doctoral candidate Aivo Olev has developed Jutusta.ee, an Estonian-language speech technology web solution that brings voice cloning capabilities within reach of ordinary users.

What can Jutusta.ee do?

The platform offers three main functionalities: converting speech recordings into written text, reading written text aloud in a human voice, and, most notably, cloning the user's own voice. According to Olev, the quality of cloning is so high that even family members and close friends cannot distinguish between the original human voice and its artificially created copy.

Estonian-language artificial intelligence

Jutusta.ee is aimed primarily at Estonian-language users, which makes it a remarkable achievement in Estonia's speech technology landscape. Estonian is a relatively small language, and developing quality speech technology solutions requires specialized expertise and datasets.

The solution inevitably raises questions about the misuse of deepfake audio-when a voice copy is so convincing that even loved ones cannot tell the difference, the question arises of how to protect people against potential deception.