Facebook parent Meta, introduced an artificial intelligence (AI)-powered chatbot that can convert a chunk of new speech from text input and a brief audio clip for context. Called Voicebox, the company said that it can perform speech generation tasks it was not specifically trained on. However, the company is not releasing the chatbot for public use.
According to Meta, despite having several use cases, there is a risk factor involved in releasing the chatbot for the masses.
“There are many exciting use cases for generative speech models, but because of the potential risks of misuse, we are not making the Voicebox model or code publicly available at this time,” Meta said.
It highlighted that the move is necessary “to strike the right balance between openness with responsibility.”
What Voicebox can do?
Voicebox is just like any other generative system for images and text but instead, it creates outputs in the shape of high-quality audio clips. Meta says that it can create outputs from scratch and even modify a sample that is fed to it that too in six languages, including English, French, German, Spanish, Polish, and Portuguese.
“The model can synthesise speech across six languages, as well as perform noise removal, content editing, style conversion, and diverse sample generation,” it said.
Use cases of Voicebox
Meta explains that Voicebox can not only create high-quality audio clips but also edit pre-recorded audio, like removing car horns or a dog barking. At the same time, it preserves the content and style of the audio.
In the future, the company says, such AI models can even give natural-sounding voices to virtual assistants and non-player characters in the metaverse or games.
Furthermore, in the case of visually impaired people, they can hear written messages from friends in their voices by using such models. Also, they can even create background music for videos.