By Stephen Nellis
(Reuters) – Nvidia (NASDAQ:) on Monday confirmed a brand new synthetic intelligence mannequin for producing music and audio that may modify voices and generate novel sounds – know-how aimed on the producers of music, movies and video video games.
Nvidia, the world’s largest provider of chips and software program used to create AI programs, mentioned it doesn’t have quick plans to publicly launch the know-how, which it calls Fugatto, brief for Foundational Generative Audio Transformer Opus 1.
It joins different applied sciences proven by startups akin to Runway and bigger gamers akin to Meta Platforms (NASDAQ:) that may generate audio or video from a textual content immediate.
Santa Clara, California-based Nvidia’s model generates sound results and music from a textual content description, together with novel sounds akin to making a trumpet bark like a canine.
What makes it completely different from different AI applied sciences is its capacity to absorb and modify present audio, for instance by taking a line performed on a piano and reworking it right into a line sung by a human voice, or by taking a spoken phrase recording and altering the accent used and the temper expressed.
“If we think about synthetic audio over the past 50 years, music sounds different now because of computers, because of synthesizers,” mentioned Bryan Catanzaro, vice chairman of utilized deep studying analysis at Nvidia. “I think that generative AI is going to bring new capabilities to music, to video games and to ordinary folks that want to create things.”
Whereas firms akin to OpenAI are negotiating with Hollywood studios over whether or not and the way the AI may very well be used within the leisure business, the connection between tech and Hollywood has turn into tense, notably after Hollywood star Scarlett Johansson accused OpenAI of imitating her voice.
Nvidia’s new mannequin was skilled on open-source information, and the corporate mentioned it’s nonetheless debating whether or not and learn how to launch it publicly.
“Any generative technology always carries some risks, because people might use that to generate things that we would prefer they don’t,” Catanzaro mentioned. “We need to be careful about that, which is why we don’t have immediate plans to release this.”
Creators of generative AI fashions have but to find out learn how to forestall abuse of the know-how akin to a consumer producing misinformation or infringing on copyrights by producing copyrighted characters.
OpenAI and Meta equally haven’t mentioned once they plan to launch to the general public their fashions that generate audio or video.
