I thought it would be fun to clone my voice using one of those generative AI tools that take a few seconds of you speaking and turn it into a reusable model. What came out was a weird version of my voice that had an American accent. Perhaps my weird mid-Atlantic half-British half-American maybe-Australian-sounding voice is too far outside expected models to clone easily? Is it possible that more "average" people are easier to fake?
@ben Same, their base models are all trained on US accents so it just applies the timbre of your voice to a US accented kind of blueprint. Won't improve until they spin up robust base models on a variety of accents and allow you to select the accent to then fine tune with the timbre of your voice.
Add comment