OpenAI is developing a new AI music generation tool that could create songs from text and audio prompts.
According to a report from The Information published October 24, 2024, the ChatGPT creator is collaborating with students from the prestigious Juilliard School to annotate musical scores and create training data for the model. The tool is envisioned to generate guitar accompaniments for vocal tracks, add music to videos, or create full compositions from simple text descriptions.
This isn’t OpenAI’s first venture into music. The company released MuseNet in 2019 and Jukebox in 2020, both experimental projects that generated music but were eventually discontinued. Jukebox was known for producing often-unpredictable audio with noticeable artifacts, making it more of a research curiosity than a practical tool.
The new project reportedly aims to be far more sophisticated, though details remain scarce. OpenAI has not announced a launch timeline or confirmed whether the tool will be standalone or integrated into existing products like ChatGPT or Sora.