🌱 Ethical by architecture
This isn't a neural model trained on scraped audio — it's a procedural composition + synthesis engine. Your chosen vibe tags become a deterministic seed that drives music theory rules (and they nudge the key, tempo and energy too), and every sound is rendered from oscillators and noise in your browser. No training data, nothing uploaded. Everything you generate is yours.
🔊 The sound engine
Detuned supersaw stacks, harmonic-layered keys, saturated 808/sub bass with glide, layered drums with crash cymbals, sidechain pumping, pre-delayed convolution reverb, tempo-synced delays, tape saturation and bus compression on every stem.
📁 Exports & troubleshooting
Clicking WAV / MP3 / MIDI / All stems performs a normal browser download. If nothing happens, the toast shows a “Not saving?” link — click it to get the system Save dialog, a new-tab view, an inline player, and a right-clickable direct link.
Important: if you are viewing this page inside an embedded preview (a chat pane, a sandboxed iframe, some online IDEs), the host can block all downloads at the browser level — no code can override that. Save the HTML file and open it directly in a normal browser tab (double-click it), and every export button will work.
🎼 Song structure
Optional intro (pads & arps, snare riser) and outro (strips back to a final chord ringing into the reverb tail) — untick the intro & outro box to jump straight into the full main groove with fills and crash hits, right through to the last bar — all scaled to the duration you choose.
🎹 MIDI export
Format-1 .mid files: tempo map + one track per stem (Drums on GM channel 10). Drop them into any DAW.