2024-03-30 16:19:00
OpenAI presented, on Friday (29), its advances with a new artificial intelligence tool that can reproduce a person’s voice with a sample of just 15 seconds. After gaining renown in the field of Artificial Intelligence with ChatGPT, DALL-E and Sora, the American startup revealed details about the “Voice Engine”.
After being trained with a person’s voice, the Voice Engine can “speak” any word, phrase or text entered by the user — even if it is not in the same native language as the individual. This means, for example, that artificial intelligence would be able to pronounce a text in Portuguese using the voice of an English-speaking person.
Voice Engine is not yet available to the public. OpenAI claims to be “committed to developing safe and broadly beneficial AI,” possibly referring to caution to prevent misuse of the technology.
According to the company, one of the potential uses of the tool is translation. Creators could use samples of their own voices to translate content — whether in a video or podcast — and reach a larger audience. Another possibility is to help people who suffer from degenerative speech conditions.
The startup presented a practical demonstration of how artificial intelligence works. An audio clip of a speech in English is provided. The Voice Engine then uses the reference voice to translate the speech into Japanese. See below.
Reference audio (English)
Audio translated by Voice Engine (Japanese)
OpenAI adds that Brown University in the United States is exploring the use of Voice Engine to help patients in the treatment of diseases that compromise speech. Experts have already restored the voice of a young woman who lost her fluency due to a brain tumor using only audio from a video recorded for a school project.
Despite the benefits that the technology could bring to medicine, the entertainment industry and other spheres, there are clear concerns about the risk of using artificial intelligence for illicit purposes, which could involve cloning the voices of people — celebrities or not — for scams. and triggering fake news.
Security
08 Mar
Tech
23 Ago
One of the cases that generated repercussions were the deepfakes used to attract victims to fake betting games on the internet. Celebrities and digital influencers had their images cloned using artificial intelligence to create false advertisements and give credibility to the scammers’ platforms.
For this reason, the Voice Engine can only be accessed by institutions authorized by OpenAI. There is no forecast of availability of this tool to the public. On a related note, OpenAI is preparing the public launch of AI Sora for 2024. It is a technology capable of generating videos through a simple text command.
1711860636
#Owner #ChatGPT #OpenAI #creates #artificial #intelligence #capable #reproducing #peoples #voices