OpenAI made a statement about the company's new technology that can clone voices.
The statement shared that the “Voice Engine” can produce speech that closely resembles the speaker’s voice using only written text and a single 15-second speech sample.
According to the statement, which says the technology in question was developed in 2022, a cautious and deliberate approach to wider release was taken due to the “potential for abuse.”
“We recognize that producing speeches that sound like the voices of the people poses serious risks, especially in an election year.” According to the statement, feedback was collected through collaboration with many international partners, including the United States .
As for security concerns, in the Voice Engine testing phase, user partners were told that it was forbidden to use the identity of individuals or organizations, the consent of the original speaker was required, and it was imperative to clearly inform the audience about the voices are generated by artificial intelligence.
Can also translate
According to the statement, Voice Engine will also be able to translate while retaining the original speaker's voice.
It was shared that content producers such as videos and podcasts will benefit from this feature and will be able to reach the world “in their own voice” while still retaining the original speaker’s “accent” in the translated audio.
Bir yanıt yazın