Alarming! AI can visualize your image from your voice

May 30, 2019, 8:29 AM EDT
(Source: Fe Ilya/flickr)
(Source: Fe Ilya/flickr)

Artificial intelligence (AI) has a knack for surprising us with its ever growing capabilities. The latest in its armor is a reasonably good faculty to infer how a person looks like by just listening to a small audio clip of their voice.

Researchers at Carnegie Mellon University created an AI algorithm, the Speech2Face, which does remarkably well in voice profiling, particularly in reconstructing someone’s face from their voice, notes Futurism.

The researchers, however, admit privacy and ethical concerns with this technology, given the potential sensitivity of facial information. As per the details of the experiment, published on GitHub page, the GAN-based framework shows considerable room for improvement despite delivering reasonable reconstruction results.