OpenAI Voice Engine can recreate human voices

OpenAI, the company behind ChatGPT, has been on a roll lately. ChatGPT’s success has of course been astounding, but the company also recently announced Sora, a system capable of creating highly realistic 60-second video clips. Today, the company announced a new system called Voice Engine, capable of recreating human voices (via technology/openai-voice-engine.html” target=”_blank”>The New York Times).

Like Sora, OpenAI doesn’t allow the public to use Voice Engine – at least not yet. For now, the company is testing the system privately with “a small group of companies.” Obviously, the reason he does this is due to the massive ethical implications of a system capable of imitating a real person’s voice.

The New York Times I was able to demo the system and share some clips, which you can listen to at the previous link. The first clip is a 16-second recording of a real man with a strong Portuguese accent. He introduces himself and says he’s making this clip to “help non-verbal individuals express themselves more fully.” The next clip is Voice Engine’s recreation of his voice saying something completely different. Yet another clip is a recreation of the man’s voice but speaking in Portuguese instead of English.

The two Voice Engine clips do not sound the same as the original clip. However, they are absolutely close enough to fool someone who knew this man’s voice into believing that he actually said these things.

What’s scary is the possibility of using a tool like this to spread misinformation. Politicians, celebrities and journalists could easily have their voices picked up by Voice Engine and then be tricked into saying whatever they wanted. With a little editing and a compelling video track, who knows what could be done?

There are also voice authentication systems used around the world for security reasons. It is very possible that Voice Engine allows users to trick these systems, putting sensitive information at risk.

“It’s a sensitive thing”

Jeff Harris, OpenAI product manager, said: “This is a sensitive issue and it’s important to get it right. » OpenAI is experimenting with watermarking systems to help differentiate real recordings from synthetic recordings. The team is also open about the ethical issues this system raises.

However, he also says Voice Engine could do a lot of good. For example, people who could once speak but lost their voices later in life might start communicating again using a facsimile of their own voice. Physicist Stephen Hawking is a famous example of someone who could have benefited from a voice service like this. Voice Engine could also preserve the voices of people who are no longer alive and work in many commercial contexts, such as in the creation of audiobooks.

OpenAI says it has no plans to publicly deploy Voice Engine yet. Like Sora, he only wants to demonstrate what he is capable of.