Meta, the company that owns Facebook, Instagram, WhatsApp, and others, announced today on Friday a revolution in the field of conversational artificial intelligence. This comes in the midst of escalating competition with other tech giants such as Google, Microsoft, and Amazon.
Meta unveiled its development of a new artificial intelligence model called Voicebox, which possesses the ability to perform speech-related tasks such as editing, sampling, and style identification. This model stands out due to its contextual learning capability, as it undergoes specific training to execute these tasks.
The technology giant believes that in the future, multipurpose artificial intelligence models like Voicebox can provide natural voices for virtual assistants and non-player characters in the metaverse.
Meta also stated that these models enable visually impaired individuals to hear written messages from friends, as they are read in their voices through artificial intelligence. They also provide content creators with new tools for easily creating and editing audio clips for videos, among other features.
Using just a two-second voice sample, Voicebox can match and utilize a person’s voice pattern to develop text-to-speech technology.
Meta also highlighted the new multipurpose artificial intelligence model’s ability to rephrase garbled speech or replace misspoken words without the need for re-recording the entire speech.
When provided with a sample of someone’s speech and a text passage in English, French, German, Spanish, Polish, or Portuguese, Voicebox can read the text in any of those languages, even if the model itself is different in the speech and text domains.
Meta hopes to leverage this capability in the future to enhance natural and authentic communication among individuals, even if they don’t speak the same languages.
With the utilization of diverse datasets, the company announced that its model has the ability to produce speech that better represents real-world communication methods in the six currently supported languages.
Meta considers Voicebox to be a significant step forward in its efforts in generative artificial intelligence. It looks forward to further exploration in the field of audio and observing how other researchers can benefit from its work.