Meta Launches AI-Powered Audiobox: Advanced Generative AI for Custom Audio Creation with Watermarking

Meta, the parent company of Facebook, has unveiled Audiobox, an AI-powered audio generator platform that allows users to create customised voices and sound effects. The platform builds on the technology behind Meta's Voicebox tool, offering improved quality and automatic watermarking to ensure responsible use. Audiobox encompasses speech, sound effects, and soundscapes, providing a range of input mechanisms for maximum control. Meta claims that the platform utilises "bespoke solvers" to generate audio over 25 times faster than previous models without sacrificing performance. This release follows Meta's announcement earlier this year of Voicebox, a generative AI tool capable of producing audio in six languages. However, due to concerns over AI-powered deepfakes, Meta vowed not to release Voicebox to the public. To address similar concerns with Audiobox, watermarking has been introduced to trace the origin of any audio created.