Meta’s reveals new text-to-speech Voicebox AI but you can’t have it

Meta has just revealed its new VoiceBox AI that delivers ChatGPT-like text-to-voice generation for its users — but you can’t have it.

It’s no secret that Mark Zuckerberg has plans for Meta to take advantage of actively advancing AI technology. They’ve even got plans to add it to Facebook and Instagram.

They’ve revealed options like an AI bot for Messenger and WhatsApp, but now Meta has revealed a new text-to-speech generation app.

With VoiceBox AI, Meta is bringing ChatGPT-like voice generation to its users — but you can’t access it just yet.

Meta reveals Voicebox text-to-speech AI

In a report from Engadget, Voicebox promises to deliver the same quality of text-to-speech generation that ChatGPT does for text and Dall-E does for images.

According to the site, Meta defines the system as “a non-autoregressive flow-matching model trained to infill speech, given audio context and text” that has been trained on over 50,000 hours of unfiltered audio from public domain audiobooks written in various languages.

With this dataset that it learned, Voicebox should be able to create a more conversational tone of speech, regardless of language.

Unsplash: Dima Solomin

The new AI from Meta is also reportedly capable of editing existing audio clips by removing noise and even replacing misspoken words.

Meta researchers explained: “A person could identify which raw segment of the speech is corrupted by noise (like a dog barking), crop it, and instruct the model to regenerate that segment.”

With the reveal of this technology, though, Meta has made it clear that AI fanatics shouldn’t get their hopes up for its release. Due to “potential risks of misuse,” Meta is not releasing the app or its source code to the public just yet.

We’ll update you as Meta continues to reveal more and more of what they have planned throughout the company. In the meantime, head over to our Tech section for more news, reviews, guides, and more.