OpenAI is developing a new application that would create music from text and audio prompts, as reported by The Information.

This application could potentially enhance existing videos with music or provide guitar accompaniment for pre-recorded vocal tracks, sources indicated. The timeline for the launch of such a tool by OpenAI remains uncertain, as does the possibility of it being offered as an independent product rather than as part of OpenAI’s ChatGPT and video application Sora.

A source shared with The Information that OpenAI is collaborating with students from the Juilliard School to annotate scores, which would serve to generate training data.

Although OpenAI has introduced generative music models in the past, they emerged before ChatGPT’s debut; more recently, the organization has been progressing with audio models concentrating on text-to-speech and speech-to-text capabilities. Other entities with generative music models include Google and Suno.

TechCrunch has contacted OpenAI for a response.