OpenAI releases new Voice Engine tool to select partners in election year

OpenAI has released a new tool that can generate a clone of anyone’s voice, based on a 15-second recorded audio, to a limited number of companies for testing.

The ChatGPT creator revealed that its Voice Engine tool has been in development since late 2022 and can be used to support tasks such as reading assistance, translation, and support for the non-verbal.

The generative AI giant said that it was also taking steps to mitigate risks in what has been termed as the global year of elections, with half the world’s population set to go to the polls this year.

Designated users of Voice Engine must agree to not use it to impersonate an individual or organisation without consent or legal right, OpenAI added.

An OpenAI blog post announcing the tool said: “Our terms with these partners require explicit and informed consent from the original speaker,” and the partners must disclose to their audience that the voices are AI-generated.

“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,” it added.

Audio deepfakes have been known to cause confusion in politics. Voters in the US state of New Hampshire, for instance, received a call with a deepfake audio of President Joe Biden advising them not to vote in the presidential primary elections earlier this year.

And, at the end of last year, the UK’s opposition leader Keir Starmer was forced to debunk a deepfake audio of him that gathered 1.5 million views on X of him ‘swearing and abusing’ party staffers.

Year of elections: a deepfake threat on politics and business

To prevent its tool from contributing to misinformation, OpenAI said that it will have a “no-go voice list” that detects and prevents the creation of voices that are too like those of prominent figures.

The firm added that it plans to encourage the phasing out of voice-based authentication as a security measure for accessing bank accounts and other sensitive information, exploring policies for individuals’ voices in AI, educating the public on deceptive AI content, and developing techniques for tracking the origin of audiovisual content for verification.

Voice Engine for good

Judging by the partners that OpenAI chose to publicly list as testing its new voice app, the tech firm appears keen to promote the societal benefits of its latest product.

The edtech firm Age of Learning, for instance, has been using the tool to generate pre-scripted voice-over content. The company is also using GPT-4 to create real-time, personalised responses to interact with students to help them study.

Deepfakes for good? How synthetic media is transforming business

Healthtech firm Dimagi is also using Voice Engine to provide counselling for breastfeeding mothers in remote areas, and to help upskill workers in their own language including Swahili or informal languages such as Sheng, which is popular in Kenya.

Communication app Livox, and the Norman Prince Neuroscience Institute at Lifespan are both using it to help those with disabilities communicate.

There were some enterprise partnerships listed too: HeyGen, a story-telling platform which is using Voice Engine to work with its enterprise customers to create human-like avatars for content production, including product marketing and sales demos.