Adobe’s New Generative AI Turns Text Prompts into Music

Adobe Research revealed a new generative AI music and audio prototype, Project Music GenAI Control, that promises to give anyone the power to create music, no matter their experience or skill. Adobe says it improves accessibility, empowerment, and democratization within music.

Project Music GenAI is the tech giant’s latest attempt to use generative artificial intelligence technology to make it easier for anyone to create art. The company’s recent and most notable efforts have come primarily in the form of generating visual content, such as Adobe Firefly, and generative expand and generative fill inside Photoshop.

However, Adobe is no stranger to using AI with audio. Beyond a terrifying and failed attempt nearly a decade ago to allow people to synthesize realistic voiceovers, Premiere began allowing users to enhance speech with AI technology last year.

But Project Music GenAI Control goes far beyond simple enhancements of real-world audio; it enables users to input a text description and receive an artificially created relevant musical creation in response. For example, people can make “inspiring film” or “intense hip-hop” music, as seen in Adobe Research’s video below.

“Explore the future of sonic creativity,” Adobe says. However, it’s not immediately evident precisely where the “creativity” comes in at the content creation level.

That said, it’s arguable that users can add a human touch through the editing they perform on the generated music clips. This is an area where Project Music GenAI separates itself from something like Meta’s AudioCraft, which produces music using text prompts but doesn’t provide the user much additional control.

While much of Adobe’s press material emphasizes the ability for anyone, regardless of talent or practice, to create any music to fit their needs, a senior research scientist at Adobe Research, Nicholas Bryan, touts the platform’s editing suite, describing what users can do to the AI-generated audio as “pixel-level control for music,” similar to what users can do with their images in Photoshop.

Users can also transform generated audio based on a selected reference melody. This can be used to adjust the tempo, structure, and repeating patterns of music. It’s also possible to adjust intensity, clip length, and generate a seamless, repeatable loop.

Adobe developed Project Music GenAI Control in collaboration with colleagues at the University of California, San Diego, and the School of Computer Science at Carnegie Mellon University in Pittsburgh. The new AI-powered audio platform is currently only a prototype and is not publicly available. Whether it will ever make it to customers in this form or not, it’s clear that Adobe is keen to put the power to synthesize audio in its users’ hands.

Image credits: Featured image created by Adobe Stock user _veiksme_