AIn recent months, a new type of deepfake has been gaining popularity, which allows you to clone any voice and create a full-fledged speech based on it. OpenAI, the manufacturer of the well-known chatbot ChatGPT, also decided to be at the forefront of these technologies and announced its new artificial intelligence Voice Engine, which converts text into speech.
All the AI needs to work is a 15-second fragment of human speech, after which the user can type any text into the program and get the finished audio recording. According to the developers, their technology was created exclusively to help people, for example, to voice books for visually impaired people or to help patients with speech problems.
However, during the testing process, OpenAI itself felt that its creepy technology was not ready for mass adoption. Creating a speech based on any voice is a tidbit for scammers who can clone the voices of famous people to influence the political life of countries. Phone thieves will not be left out either, who, with the help of such AI, will be able to pretend to be relatives of victims in order to swindle money.
That is why the company decided to slow down the spread of the new technology, and tests only with selected partners. There are a number of requirements for testers | for example, they are required to report that the audio track was generated by AI, as well as to obtain consent for cloning from the voice owner. Developers are now integrating Voice Engine’s capabilities into their own ChatGPT-4 language model to create entertaining and educational content.
