Meta introduces a new AI tool that can imitate the voices of friends and family. (Image: Meta)
Meta has introduced a new AI tool called Voicebox. The tool can convert text to speech using anyone’s voice. According to the company, the feature should primarily help people with visual impairments. For such people, the tool could, for example, read out text messages – in the voice of the sender. However, Meta admits that there is a risk of improper use of the voice box.
Voicebox has three core functions
According to Meta submits two seconds long audio clip so that the voice box can reliably imitate a voice. In the presentation video, the company shows how the AI ​​tool can render a sentence in multiple speech styles and voices.
In addition, the new AI tool dominates six different languages – namely: English, German, French, Spanish, Polish, Portuguese.
It is capable of reading texts in any of these six languages in any voice to play back. Meta wants to enable more people to communicate with each other in a “natural” way, even if they don’t speak the same language.
Otherwise, the voice box is also able to free speech recordings from background noise, such as dog barking or car horns. Meta describes the function, like a »Eraser for Audio Editing«.
What are the areas of application of Voicebox?
The Internet giant will not release the new Voicebox app or the source code to the public. Although they find the new technology particularly exciting and there are many different areas of application, they overestimate the risk of misuse.
Instead, the company plans to use the text-to-speech generator for NPCs in the Metaverse, voice assistants, or even in voice prostheses.
Meta’s new tool isn’t the first in this field. It is therefore somewhat surprising that the company has classified its own as too risky. What do you think of AI apps like the Meta Voicebox? Are they an important technological step forward or are they too risky to release to the public? Tell us what you think about it in the comments!