TORONTO — A Toronto lab has found a way to tell whether man or machine is behind most audio clips.

Staff at Klick Health's research arm say the method is based on vocal biomarkers, which are features embedded in voices that tell us things about the speaker’s health or physiology.

The five biomarkers Klick focused on include pauses when someone is speaking, takes a breath or grasps for words, which artificial intelligence-generated audio doesn't do.

These biomarkers have given Klick an 80 per cent success rate when it tries to identify deep fakes, which are A-I-produced video, audio clips or photos that look extremely real.

The European Union’s law enforcement agency Europol recently predicted as much as 90 per cent of online content may be synthetically generated by 2026.

Klick is hopeful its research can help advance the fight against deep fakes.

The Canadian Press