It is now possible to clone voices and OpenAI has a new solution. The platform is called Voice Engine and can generate voice messages from a 15-second audio with your voice. It is important to highlight that access is currently limited.
According to The Verge, the Voice Engine can read text instructions and turn them into audio. Currently, OpenAI resources are accessed by companies such as Age of Learning, HeyGen, Dimagi, Livox and Lifespan.
Regarding the development of the Voice Engine, a member of OpenAI explained how to develop the model. In the interview given to TechCrunch, Jeff Harris assured that it is “a combination of licensed and publicly available data”.
Remember that voice cloning techniques have been a controversial subject. So much so that the United States government has tried to eliminate the use of this resource. As The Verge reports, several Americans even received calls with Joe Biden’s voice, cloned by Artificial Intelligence (AI).
Open AI ensures that partners have agreed to the usage policy
Regarding this topic, OpenAI guarantees that everything has been fulfilled. The brand states that the partners agreed to the usage policy, which abhors any type of appropriation of other identities.
Another element that the company introduced were watermarks. Through these, OpenAI intends to track what is generated by the Voice Engine. This way, there is more control over how the tool is used.
Either way, OpenAI suggests some development steps to reduce risks as much as possible. According to the company, the goal is “a cautious and informed approach to a wider rollout due to the potential for misuse of synthetic voices” (via Exame).