AI Voicebox: Multilingual Speech Generation

Voicebox is a cutting-edge speech generative model developed by Meta that uses a non-autoregressive flow matching model to generate multilingual speech. The AI has been trained on a large scale of data to solve a text-guided speech infilling task, outperforming single-purpose AI models across various speech tasks through in-context learning. It can synthesize speech in six different languages, remove transient noise, edit audio content, transfer audio style within and across languages, and generate diverse speech samples. It also generates speech up to 20 times faster than other state-of-the-art auto-regressive models.