// Cantis landing page sections — music-themed, with real-app-mirroring demo
// =================== Helpers ===================
const Mono = ({ children, dim, className = '', style = {} }) => (
{children}
);
const Arrow = ({ size = 14 }) => (
);
// =================== Top Nav ===================
function TopNav({ accent }) {
return (
Cantis runs ACE-Step v1.5 natively on your Mac. Prompt a track, edit lyrics,
tune the seed — everything stays on-device. No queue, no cloud, no upload.
Generate
music that
actually sounds
like yours.
Built in Swift, accelerated by MLX. A 4-minute track renders in under 30 seconds on M2 Pro. No Python, no Docker, no GPU required.
Real-time playback with waveform visualization and FFT spectrum analysis. Watch low-end, mids, and air shape themselves bar by bar.
text2music, cover, repaint, extract. Drop in reference audio for the last three; the model rewrites, restyles, or pulls a part out.
Verse / chorus / bridge with ISO-639 language hints. Layer genre, instrument, and mood tags to steer the model.
Cantis ships three DiT variants downloaded on first launch from HuggingFace. Swap them per-render — fast iteration, polished final, or balanced base.
Steps, schedule shift, and CFG scale — exposed where they matter.
Drag-and-drop source audio for cover, repaint, or extract modes — and the model takes it from there.
Save reusable generation configs. Browse, search, and favorite past tracks — every render kept locally with its prompt and parameters.
WAV, AAC (.m4a), or ALAC (.m4a) via Apple's encoder. Drag the file straight into Logic, Ableton, or Reaper.
{it.body}
What you need to run Cantis. What you get when you do.
| {k} | {v} |