OpenAI launched voice capabilities in ChatGPT final September. Now, the corporate is previewing a mannequin known as Voice Engine, which may use a single 15-second audio clip and textual content immediate to generate longer audio. OpenAI boasts that Voice Engine produces life-like voices with inflection and tone, slightly than a robotic drone.

In response to OpenAI’s weblog put up, Voice Engine was first developed in 2022 to energy ChatGPT’s Learn Aloud function in addition to text-to-speech. Since then, OpenAI has examined Voice Engine in a lot of completely different eventualities: for kids and non-readers; for non-verbal individuals and individuals who have in any other case misplaced their voice; and translation. For every of those circumstances, OpenAI has partnered with firms in these respective areas.

OpenAI spends a bit of the weblog put up assuring that Voice Engine is constructed safely — and whereas it would not point out the notorious robocall of Joe Biden explicitly, it is implied: “We acknowledge that producing speech that resembles individuals’s voices has critical dangers, that are particularly prime of thoughts in an election yr,” the put up states. “We’re partaking with U.S. and worldwide companions from throughout authorities, media, leisure, training, civil society and past to make sure we’re incorporating their suggestions as we construct.”

The Biden robocall was possible made with software program by ElevenLabs, not an OpenAI product, however Voice Engine may maintain the identical capabilities. On the finish of the put up, OpenAI states that as a result of its “strategy to AI security and [their] voluntary commitments” (committing to security, safety, and belief), it isn’t releasing Voice Engine extensively proper now.

“We hope this preview of Voice Engine each underscores its potential and likewise motivates the necessity to bolster societal resilience in opposition to the challenges introduced by ever extra convincing generative fashions,” the put up states. OpenAI then requires a phase-out of voice-authentication as a safety measure; insurance policies to guard using individuals’s voices in AI; training; and extra know-how to establish inauthentic voices — all due to its personal know-how.

Learn extra and take heed to examples of Voice Engine in OpenAI’s weblog put up, Navigating the Challenges and Alternatives of Artificial Voices.