Home Tech Meta has developed a generative artificial intelligence model for text-to-speech

Meta has developed a generative artificial intelligence model for text-to-speech

0
Meta has developed a generative artificial intelligence model for text-to-speech

[ad_1]

Meta has developed a generative artificial intelligence model for text-to-speech

Meta introduced a generative text-to-speech model Voicebox. According to the developers, the algorithm will do for oral speech what ChatGPT and DALL-E did for text and images.

What is known

Similar to generative systems for text and images, Voicebox can create output from scratch, transform styles, and modify the provided template. The system was trained on 50,000 hours of recorded speech and transcripts of public domain audiobooks in English, French, Spanish, German, Polish and Portuguese.

As a result, Voicebox is able to edit clips, eliminate noise, and replace mispronounced words.

“A person can determine which raw segment of speech is damaged by noise (such as a barking dog), clip it, and instruct the model to regenerate that segment,” the researchers said.

Voicebox can also play speech over a two-second passage, transfer cross-language style, and create a variety of samples for synthetic data sets.

When to expect

Meta did not publish the source code of the model. The developers referred to “potential risks of misuse”despite “many interesting use cases for generative speech models”.

Source: Meta.



[ad_2]

Source link

gagadget.com

LEAVE A REPLY

Please enter your comment!
Please enter your name here