Write the lyrics. Pick a style. Hear the song.
Alunan is a planned free, open-source desktop app that generates music from your lyrics and a musical style, entirely on your own computer. No terminal, no separate installs, no cloud service.
What Alunan is being built to do
These are the goals in the project's feature specification. None of them is available in a released app yet.
Lyrics and style to a song
Enter lyrics with section tags and describe the style you want. Alunan generates one complete song you can play and save as WAV.
Runs locally
Generation happens on your machine. Lyrics, styles and audio are never uploaded for inference, and there is no telemetry in the app.
Models download themselves
The first song downloads the music model it needs, plus NVIDIA's GPU components on PCs that use them, verifies everything and carries on generating. After that it works offline.
Nothing else to install
One installer, and the app fetches anything else it needs by itself. No Python, package managers, toolkits or servers to set up.
Honest progress and Stop
Real download bytes, clear stages from checking to saving, and a Stop that actually stops the whole operation while keeping reusable downloads.
Free and open source
Alunan's source is Apache-2.0. Music models keep their own licenses, which the app shows separately.
Native on every desktop
Each platform gets its own native interface built from the same specification. Supported hardware will be published only after it has been measured.
| Platform | Interface | Status |
|---|---|---|
| Windows | WinForms | Planned — engine feasibility testing |
| Linux | GTK 4 | Planned — needs native test hardware |
| macOS (Apple Silicon) | Swift | Planned — needs Mac test hardware |
Where it stands
Alunan is in Phase 0: proving that a complete song can be generated locally on each platform and packaged without extra installs. The app itself has not been built yet, and there are no downloads.
- Prove the music engine and models work locally on Windows, Linux and macOS.
- Build the native app shells and shared data formats.
- Add model downloading and the generation engine.
- Deliver the first-song flow, then the full app.
- Test reliability on every platform and release installers.
Progress and measurements are recorded in the implementation plan and the Phase 0 findings.