Meta open sources models for generating sounds and music

Meta's announcement of Audiocraft, a framework for generating high-quality audio and music using generative AI models. The framework includes three models: MusicGen, AudioGen, and EnCodec. MusicGen can be trained on user data but raises ethical and legal concerns. AudioGen generates environmental sounds and sound effects. EnCodec is an improved model for compressing and reconstructing audio. Meta acknowledges the potential for misuse and ethical questions but plans to continue improving generative audio models.