The adage “don’t believe everything you see on the internet” might need an update, as you may not be able to trust what you hear in the near future. A new online tool called Koe Recast aims to allow users to alter their voice utilizing AI algorithms. Koe Recast doesn’t just alter the pitch or timbre of your voice, it harnesses AI to completely transform your speech until it sounds like it is coming from someone else entirely, all while retaining your tone.
The Koe website features several demonstrations using a short clip of Mark Zuckerberg speaking about augmented reality as its base input. From there, outputs are shown using Kore Recast to have Zuckerberg deliver the same words in the voice of a male narrator, a female voice, and finally a high-pitched woman that would be typically heard in cartoons or anime.
Still in its alpha stages, Koe Recast is being developed entirely by one person, a Texas-based engineer named Asara Near. As reported by Ars Technica, Near hopes Koe Recast will be implemented as a desktop app that users could use with platforms such as Zoom and Discord. The fact that such impressive technology can be developed by just one person may forecast what incredible tech might be on the horizon, especially by large teams with larger funding.
In their talk with Ars Technica, Near states that it is unclear if Koe Recast will eventually allow users to use the voices of celebrities or well-known persons. On the ethics of technology like Recast, Near says, “As with any technology, it’s possible for there to be both positives and negatives, but I think the vast majority of humanity consists of wonderful people and will benefit greatly from this.” Additionally, Near noted that there are clauses in Recast’s Terms of Service that prohibit illegal or hateful usage.
While Recast is still in developmental stages, Near hopes to pursue commercial options, telling Ars Technica , “We’re exploring some monetization strategies,” while also expanding that an open-source release is not completely out of the question.