OpenAI Unveils Groundbreaking 'Voice Engine' for Text-to-Speech Technology
OpenAI has created a new AI-based tool called Voice Engine. It allows to read text in the voice of another person. This is stated in the company's blog.
The new Voice Engine tool has several options for use. In particular, it can very realistically voice texts with the voice of any person. A 15-second sound is enough to train it.
In addition, this tool can convey not only the direct sound of a person's voice, its timbre, but also the peculiarities of a particular speech. During translation, Voice Engine preserves the native accent of the original speaker: for example, if the original source was a French voice, the English audio will have a French accent.
According to the developers, the tool can help record and translate podcasts and videos;
people who cannot read for some reason; those who have problems with voice restoration due to various speech disorders; children in education; and mute people communicating by voicing their text requests.
Voice Engine was launched in 2022, but is not yet available. The company believes that the tool can be used unfairly on a massive scale and will become a part of the creation of realistic diplomatic facial expressions.
Therefore, the company first wants to receive feedback on possible dangers and discuss ethical issues.
At the beginning of March, OpenAI introduced another new feature for ChatGPT: Read Aloud. It allows users to listen to answers to queries in one of five voice options. The feature supports 37 languages and is available in iOS and Android apps.