Full music player
Play, pause, seek, queue, MiniPlayer. The player always works, with or without the visual engine.
THE PLAYER
A real music player under the visuals.
All visuals run on top of a full player — not a demo. Library, EQ and playlists work with zero models installed.
Audio-reactive visual engine
A diffusion model renders a real-time visual field from the live audio signal — natively via ONNX Runtime.
10-band EQ + replay gain
Sound shaping in the decode path: normalize loudness and sculpt the mix before it reaches your ears.
Library & smart playlists
Folder scanner, album art, metadata and smart playlists backed by a bundled SQLite database.
Device sync
A Rockbox / USB-MSC sync wizard keeps portable players in step with your library and playlists.
Optional by design
Visual engine models download on first use — non-modal, background, cancelable. Never a gate.
THE RULE
The player is never gated behind the models.
Boot straight into a usable player. The visual engine is an extra you opt into — surfaced non-modally, downloaded on first use, always optional.
ENGINE
On your hardware, end to end.
No cloud round-trip: audio in → decode → EQ → inference → canvas, all local.
ENG-01
Diffusion visual engine
Streams the live audio signal through a diffusion model to produce the visual field, frame after frame, on-device.
ENG-02
First-use download
Models are fetched the first time you use them, in the background, with progress shown in-app. Cancel anytime.
ENG-03
Zero-lock-in
No model, no problem. Core features are model-free by construction — the visual field is an opt-in extra.
- Rust
- ONNX Runtime
- Tauri v2
- Bun / TypeScript
- WebGL2
- SQLite
GET IT
Let the music
dream in color.
Free on Windows. Music player first, visuals after — and only when you want them.
Download for WindowsWindows 10/11 · Tauri v2 · MIT license