Meta released MusicGen, an open-source AI music generator that's described as "ChatGPT for audio

 

Meta released MusicGen, an open-source AI music generator that's described as "ChatGPT for audio."  Trained on 20,000 hours of licensed music and more, MusicGen can create new songs from text or melody prompts. MusicGen allows users to describe their desired music style, optionally provide a tune, and click "generate" for about 12 seconds of audio.  While it can't produce vocals, it specializes in instrumentals, though the generations tend to fall short of professional quality. A demo is now available on Facebook's Hugging Face AI site. The release comes about a month after Google made its text-to-music AI, MusicLM, available for testing. The model generates a 20-second music sample based on a text prompt, with some samples including "wawawa" vocals. The global market for generative AI in music was valued at $229M in 2022. By 2032, it is projected to generate revenue of over $2.6B.