In May, caller identification service Truecaller released an AI voice scanner that the company said could “alert users in real time during a call if it detects that the voice it's listening to is AI-generated.” The scanner is an AI model trained to distinguish between human voices and AI-synthesized voices, allowing users to identify when they're receiving AI-generated scam/spam calls. There was a lack of transparency about how the scanner is trained and specifically specifying what it scans calls to look for as identifiers of AI-generated voices.
Speaking to journalist Kamya Pandey of MediaNama, Raphael Mimoun and Roy Festinger from the Truecaller team said that the model currently only supports English voices and the company is working on adding Hindi support as well.
In this interview, we take a closer look at the model's capabilities, limitations, and the overall challenges in detecting audio deepfakes.
Check out the full interview here:
Read also:
Stay on top of tech news: Get MediaNama's daily newsletter with the day's top stories delivered to your inbox by 9 a.m. Click here to sign up now.
