I Lost My Voice, So I Cloned It With Generative AI




I’m a YouTuber, and as a writer I also often use my voice to dictate blog posts. It’s a great way to increase my productivity, and in video content, my voice becomes part of my brand. I sing in a choir, and I’ve been working on a project over the last month that required me to produce over 250 YouTube videos in 40 days. I used my voice a lot, and it wore out. I started to suffer from laryngitis. After an especially vocally-challenging weekend, I could barely speak. That was a problem, both in terms of my own personal life and in terms of my ability to get my work done. I wanted to stay productive and keep making videos, but I didn’t want to strain my voice further. So I tried something different; I used today’s powerful generative AI to clone my voice.

Computer voices used to sound synthetic and awful. To a large extent, the computer voices we hear on a day-to-day basis still do. No one would accuse Siri or Alexa of sounding realistic. You can tell they’re computers from the way their voices sound, and also from the way they speak — monotonously, and with very little emotion or varying emphasis. That’s a shame because today’s generative AI is capable of generating incredibly realistic computer voices. Generative AI can also easily clone a voice using a remarkably short sample.