Adobe’s latest AI experiment generates music from text

This week, Adobe unveiled an experimental audio AI tool to join Photoshop’s image-based AI. Described by the company as an “early-stage generative AI tool for generating and editing music,” Adobe’s Project Music GenAI Control can create music (and other audio) from text prompts, which it can then refine within the same interface.

Adobe sees Firefly-based technology as a creative ally that – unlike generative audio experiments like Google’s MusicLM – goes a step further and skips the hassle of moving output to external apps like Pro Tools, Logic Pro or GarageBand for editing. “Instead of manually cutting existing music to make an intro, outro, and background audio, Project Music GenAI Control can help users create exactly the tracks they need—solving end-to-end workflow pain points,” Adobe wrote in a blog post message.

The company suggests starting by entering text such as “power rock,” “happy dance,” or “sad jazz” as a base. From there, you can enter more prompts to adjust its tempo, structure, and repetition, increase its intensity, extend its length, remix entire sections, or create loops. The company says it can even transform audio based on a reference melody.

Adobe says the resulting music is safe for commercial use. It also integrates its content credentials (“nutrition labels” for generated content), an attempt to be transparent about the AI-assisted nature of your masterpiece.

“One of the exciting things about these new tools is that they’re not just about generating audio – they’re taking it to the level of Photoshop, giving creators the same kind of deep control to shape, tune and edit their audio. It’s a kind of pixel-level control for music,” wrote Adobe Research scientist Nicholas Bryan.

The project is a collaboration with the University of California, San Diego and the School of Computer Science at Carnegie Mellon University. Adobe’s announcement emphasized the experimental nature of Project Music GenAI Control. (It didn’t reveal much of its interface in the video above, suggesting it may not yet have a UI.) So you may have to wait a while before the feature (probably) makes its way into the Creative Cloud suite of Adobe.