Edgar Cervantes/Android Authority
long story short
- OpenAI has announced the launch of Speech Engine, a new artificial intelligence system capable of reproducing the human voice.
- The company is testing the product with “a small group of businesses.”
- OpenAI is keeping it secret for now to examine potential (and apparent) dangers.
OpenAI, the company behind ChatGPT, has been making good progress lately. ChatGPT’s success is staggering, of course, but the company also recently released Sora, a system capable of creating 60-second video clips that look remarkably realistic.Now, the company has announced a new system called Speech Engine that can recreate the human voice (via New York Times).
Like Sora, OpenAI doesn’t allow public use of the Voice Engine—at least not yet. The company is currently testing the system privately with “a small group of businesses.” Apparently, the reason it does this is because of the huge ethical implications of a system that can imitate a real human voice.
New York Times The system was demonstrated and some clips shared, which you can hear at the link ahead. The first clip is a 16-second live recording with a thick Portuguese accent. He introduced himself and said the video was made to “help non-verbal individuals express themselves more fully.” The next clip is a speech engine recreation of his voice, saying something completely different. Another clip is a recreation of the man’s voice, but speaking in Portuguese instead of English.
Both speech engine clips sound different from the original clip. However, they were definitely close enough that someone who knew the man’s voice would likely be fooled into thinking he actually said those words.
What’s scary is the potential for using tools like this to spread misinformation. Politicians, celebrities, and journalists can easily have their voices absorbed by a speech engine and let them say whatever they want. With some editing and a convincing video track, who knows what could be done?
Voice authentication systems are also used around the world for security reasons. Speech engines could very well allow people to trick these systems, putting sensitive information at risk.
“This is a sensitive matter”
“This is a sensitive thing and it’s important to get it right,” said Jeff Harris, product manager at OpenAI. OpenAI is experimenting with a watermarking system to help differentiate between actual and synthetic recordings. The team is also open to the ethical issues raised by the system.
However, it also believes that speech engines can do a lot of good things. For example, someone who once could speak but later lost their voice can begin to communicate again using a facsimile of their own voice. Physicist Stephen Hawking is a famous example of someone who could benefit from such a voice service. Speech engines can also preserve the voices of deceased people and can also be used in many business contexts, such as creating audiobooks.
OpenAI said it currently has no plans to publicly launch Voice Engine. Like Sora, it just wants to show what it can do.