IE 11 Not Supported

For optimal browsing, we recommend Chrome, Firefox or Safari browsers.

How many seconds does OpenAI need to clone your voice?

Answer: Fifteen.

Silhouette of a human face made from light blue dots and connected lines. A soundwave is coming from the mouth to indicate speech, also in light blue. Dark blue background.
The latest offering from software company OpenAI is equal parts impressive and concerning. Called Voice Engine, it pretty much does exactly what its name suggests by accurately replicating the sound of someone’s voice.

The tool only needs to listen to a 15-second example of someone speaking in order to do so. Voice Engine has been in the works since 2022 and was built on OpenAI’s pre-existing text-to-speech API, which it has used for its popular AI chatbot ChatGPT. The company envisions Voice Engine as a helpful tool for language translation, reading assistance and assisting people who have speech conditions.

They are also aware of the technology’s potential for misuse, which is why it is not yet publicly available. OpenAI is currently taking feedback from “U.S. and international partners from across government, media, entertainment, education, civil society and beyond” in order to improve the tool’s safeguards against nefarious uses.