OpenAI unveils voice-cloning tool but delays release on safety grounds

ChatGPT maker OpenAI has unveiled a voice-cloning tool but has no plans to release it yet as it tries to iron out safety concerns surrounding the latest addition to its generative artificial intelligence arsenal. The service, Voice Engine, has been in development since late 2022 – about the same time when ChatGPT was launched – and uses text input and a 15-second audio sample to generate “natural-sounding speech that closely resembles the original speaker”, California-based OpenAI said in a blog post on Friday . It is aimed at reading assistance, translation, supporting people who are non-verbal and helping patients with degenerative speech conditions recover their voice, as well as being a service to help hasten community services in remote areas, it said. Voice Engine joins OpenAI's suite of generative AI platforms, including Dall-E for images, Sora for video and ChatGPT, which began the generative AI revolution. Early testers will be able to preview Voice Engine, with no plans to go public “at this time”, OpenAI said, in line with the company's “approach to AI safety and our voluntary commitment”. “We are taking a cautious and informed approach to a broader release due to the potential for synthetic voice misuse. We hope to start a dialogue on the responsible deployment of synthetic voices and how society can adapt to these new capabilities,” OpenAI said. The company did not provide a timeline for Voice Engine's general release. “Based on these conversations and the results of these small-scale tests, we will make a more informed decision about whether and how to deploy this technology at scale,” it said. Voice cloning has long been used in industries and supports tasks such as customer support and engagement, as well as contribute to cost and operational efficiencies. The global voice cloning market is projected to hit nearly $9.3 billion by 2030, from an estimated $1.45 billion in 2022, growing at a compound annual rate of more than 26 per cent, latest data from Grand View Research shows . However, with the advent of generative AI, risks have also come to the fore, which include misinformation, fakery and bias, among others. OpenAI, in particular, acknowledged that these risks are heightened in 2024, being an election year in the US, where a rematch is expected between incumbent President Joe Biden and Donald Trump, who engaged in a highly-polarising, social media-fuelled battle in 2020. “We recognise that generating speech that resembles people's voices has serious risks … we are engaging with US and international partners from across government, media, entertainment, education, civil society and beyond to ensure we are incorporating their feedback as we build,” OpenAI said. OpenAI said its Voice Engine partners have agreed to its use policies, which prohibit the impersonation of another individual or organisation without consent or legal right, and require “explicit and informed consent” from the original speaker. “We don’t allow developers to build ways for individual users to create their own voices. Partners must also clearly disclose to their audience that the voices they're hearing are AI-generated,” the company said. Meanwhile, OpenAI and its biggest backer, Microsoft, are in talks to develop a data centre powered by a supercomputer built from millions of specialised chips, The Information reported on Friday, quoting sources. The supercomputer, said to be named Stargate, is expected to cost as much as $100 billion and would be based in the US, it said. Microsoft, which reportedly invested $10 billion into OpenAI in January 2023, will be most likely to finance the project, which would be about 100 times more costly compared to present data centre projects, the report said. OpenAI and Microsoft have not commented on the report.

OpenAI unveils voice-cloning tool but delays release on safety grounds

Voice Engine, available to some early testers, needs a 15-second voice sample to create a natural-sounding clip