Scroll down for the explanation.
If you guessed it was fake, you’re correct! This clip was generated using AI methods. The green box shown on the sound wave captures an intake of breath. Listening for breathing is one tool you can use when trying to discern between real and fake speech, but hearing a breath doesn’t always mean that a sample is real! As you can see in this clip, audio deepfakes are able to replicate breaths. It’s important to listen holistically, and keep in mind that human speech is made up of a constellation of different characteristics and features. For that reason, we can’t rely on just one to tell us definitively that a speech sample is real or fake.