Meta unified audio foundation model controlling spoken voice, sound effects, and acoustic inpainting
Free online access on Meta demo platform to test voice generation, prompt cloning, and sound infilling
Audiobox is Meta breakthrough unified audio generation foundation model. While voice synthesis, Foley sound effects, and audio editing historically required disparate tools, Audiobox merges speech and soundscapes into a single architecture. It supports shared text and voice prompts to guide styles and acoustics, while introducing revolutionary "Audio Infilling"—allowing users to highlight a section of an audio waveform and replace, remove, or modify sound events using text descriptions. It also delivers zero-shot voice likeness matching.