AI tool can tell real voices from fake ones
A Panjab University team has built an AI tool that can tell real voices from AI-cloned ones, aiming to fight the rise of deepfake audio.
Created by master's student Pratibha and guided by Professor Kewal Krishan, the software was trained on 100 voice clips—half real, half fake—and checks for patterns in tone and speech.
The tool already hits 80% accuracy
With voice cloning tech getting better (and scarier), this tool could help stop criminals from using fake voices in fraud or threats.
It already hits 80% accuracy and just got copyright protection from the Indian government.
The university plans to expand its use and partner with industry, while Pratibha hopes to boost its accuracy further during her PhD research.
University seeks industry partners to roll out software
Panjab University will keep rights to the software and is looking for partners to roll it out more widely—a timely move as voice fakes become a bigger problem online.