Meta Platforms is releasing AudioCraft, a generative AI framework that creates “high-quality,” “realistic” audio and music from text prompts. AudioCraft consists of three models: MusicGen, AudioGen and EnCodec, all of which Meta announced it is open-sourcing. Released in June, MusicGen was trained on Meta-owned and licensed music, and generates music from text prompts, while AudioGen, which was trained on public domain samples, generates sound effects (like honking horns and barking dogs) from text prompts. The EnCodec decoder allows “higher quality music generation with fewer artifacts,” according to Meta.

The company also announced it is releasing its pre-trained AudioGen models to let people generate their own environmental sounds and sound effects and will share all of the AudioCraft model weights and code.

“Imagine a professional musician being able to explore new compositions without having to play a single note on an instrument, or a small business owner adding a soundtrack to their latest video ad on Instagram with ease,” Facebook-parent Meta writes of “the promise” of AudioCraft.

“The day is fast approaching when generative AI won’t only write and create images in a convincingly human-like style, but compose music and sounds that pass for a professional’s work, too,” TechCrunch writes of AudioCraft, explaining that the framework simplifies functionalities previously pioneered by apps like Riffusion, Dance Diffusion and OpenAI’s Jukebox.

Since AudioCraft’s code is open source, it “provides a collection of sound and music generators plus compression algorithms that can be used to create and encode songs and audio without having to switch between different codebases,” reports TechCrunch.

But because Meta is releasing the training code, users will be able to train their own models using self-selected music samples, something “that could raise major ethical and legal issues,” TechCrunch points out.

“People can easily extend our models and adapt them to their use cases for research,” Meta AI writes in a highly technical blog post that suggests “there are nearly limitless possibilities once you give people access to the models to tune them to their needs.”