Meta Announces New Artificial Intelligence That Can Clone Your Voice With 2 Seconds Recording: Voicebox

131
Meta Announces New Artificial Intelligence That Can Clone Your Voice With 2 Seconds Recording: Voicebox

With the development of artificial intelligence technologies, we were able to clone our own or someone else’s voice and make sentences with this voice. While platforms such as ElevenLabs and Uberduck were at the top in this regard, a surprise move came from Meta today.

Meta has announced its new “Voicebox” artificial intelligence that allows you to generate speeches with artificial intelligence. Voicebox allows you to clone your voice, just like what we just mentioned, and to speak the text you write with this voice. Of course, the main event lies in the vocalization of the text, just like a human.

Introducing Meta’s voice technology Voicebox:

  • Speech style cloning with Voicebox.

Voicebox, shared by Meta as “a groundbreaking invention for productive artificial intelligence in speech”, will not only have functions such as imitating voice and reading what is written. AI will do much more than make you speak different languages, including:

“Voicebox can produce high-quality audio clips while preserving the content and style of the audio, and can edit pre-recorded sounds such as raising car horns or dog barking. The model is also multilingual and can generate speech in six languages.”

Voicebox will be able to voice a content in English, French, German, Spanish, Polish or Portuguese by taking your voice.

Moreover, for voiceover in different languages, you will not need to present text or voice in that language. The AI ​​will be able to translate a French voice or text into English or any other supported language.

In just 2 seconds, audio can be cloned:

While today’s audio cloning platforms require at least 5 minutes of recordings for the cloning process, Meta has truly broken ground here. The company stated that Voicebox can learn the voice style with just a 2-second recording and transfer this style to the voiceover.

  • Sound editing work.

In addition to all these, the words you mispronounced without noticing while recording the voice can be edited later via artificial intelligence with Voicebox.

The company has published its research paper on Voicebox. It also published the demo page where users can hear the first sounds of artificial intelligence. However, artificial intelligence is not yet available due to the potential for abuse. For now, it will only be open to scientific studies.