As if the website that generates an endless supply of fake faces wasn’t creepy enough, we now have technology that generates life-like speech using only text inputs.
While this type of software has been around for quite sometime the technology behind it is vastly improving every year. Audio clips from the video below of Joe Rogan were created using artificial intelligence by Dessa Machine Learning Engineers, Hashiam Kadhim, Joseph Palermo, and Rayhane Mama.
According to Dessa, “The Engineers used artificial intelligence to recreate Joe Rogan’s voice, generating the most human-like voice synthesis to date. The audio you are listening to is 100% generated from the artificial intelligence model. The model even learned to generate breaths and mouthing sounds where it sees fit in order to make the speech sound most natural.”
The artificial voice below is basically indistinguishable from Joe Rogan’s real voice.
Listen for yourself:
They also created a Fake Joe Rogan website with audio clips to see if you can beat their AI model. To understand how they developed the technology and to discuss the ethical implications of the work the company is doing, you can read their article over on Medium.