From prompt to track: Stable Audio 3.0 comes to Ableton and Logic

Stable Audio 3.0 is coming to DAWs through an AU/VST3 plugin, while its web app becomes an editable, iterative multitrack session.

A generated track rarely lands exactly right on the first attempt. The rhythm works, but the low end takes up too much space. The build arrives too early, an instrument feels out of place, or the structure needs further adjustment. Stable Audio 3.0’s new production experience focuses on this middle stage, between the first generation and a track that is ready to use.

Stability AI is introducing two complementary ways to access its audio model. A new Stable Audio plugin brings generation directly into music production software, while the web app becomes a multitrack environment where users can refine, direct, and assemble a creation without starting over.

The plugin works as a software instrument. In Ableton Live, it can be added to an instrument track from the plugin folder. In Logic Pro, it loads in an instrument slot. Generated audio then becomes part of the current session and can be cut, moved, processed, automated, or mixed alongside the rest of the project.

Generation can be synchronized with the session BPM, reducing the need for tempo adjustments after import. Duration is also adjustable, ranging from a short section to a complete six-minute track. Multiple versions can be saved in a playlist, allowing producers to compare takes before deciding which one stays in the arrangement.

This integration reduces the back-and-forth between a browser, downloaded files, and the DAW. It does not replace the production tools available in Ableton or Logic, but places Stable Audio alongside the instruments, recordings, and effects already used to build a track.

The plugin beta is available for macOS in AU and VST3 formats, with support for both Intel and Apple Silicon Macs. Stability AI lists Logic Pro and Ableton Live among the compatible applications but has not yet provided a comprehensive list. The announcement does not mention a Windows version or an AAX release for Pro Tools.

The new browser-based workspace takes a different approach. After generating an initial track from a prompt, users can request changes in natural language, such as reducing the low end, delaying a build, changing the instrumentation, or exploring another musical direction. The system retains the session and applies new instructions without forcing the user to return to the starting point.

Audio-to-audio can also turn an existing sound into variations across different genres, moods, or arrangements. The application organizes the resulting elements across separate tracks, with controls for volume, pan, mute, and solo. Effects can be applied to an individual track or to the master output.

Selected parts can be exported separately, while a bounce combines the session into a final mix. This structure brings the application closer to a music preproduction environment without trying to reproduce the full depth of a DAW. The browser is used to develop an idea and explore variations, while the plugin or exports allow the work to continue in a more complete production session.

Both interfaces are powered by Stable Audio 3.0, a model family that can generate music and sound effects in 44.1 kHz stereo, with some versions supporting tracks longer than six minutes. It handles text-to-audio generation, transformations from uploaded audio, edits to selected sections, and track continuation.

Stability AI says the family was trained on licensed content and that users retain ownership of their outputs. Commercial usage rights still depend on the selected plan and license: the Personal license covers noncommercial projects, the Creator license supports commercial use by individuals, and the Enterprise license is intended for larger organizations.

The plugin and redesigned web workspace remain in early beta. Some features are still experimental, and Stability AI has not detailed plugin credit usage, DAW generation times, or a release schedule for Windows. The direction is nevertheless clear: generation is becoming the start of a production session rather than its final result.