Meta has announced a new generative AI model named “Voicebox,” capable of producing high-quality speech in multiple languages and voices. The model can mimic tone and style from a short sample of audio and can even repair corrupted recordings, making it a significant breakthrough in voice synthesis.
Although Meta has not released Voicebox publicly due to concerns about potential misuse, the company has demonstrated its capabilities in speech generation, translation, and editing. This development represents Meta’s continued push into AI research and expands the use cases for synthetic voice technology in future applications like virtual assistants, accessibility tools, and content creation.