Cursor Fugue by h74611264

Cursor Fugue is an offline multi-take movement studio that treats cursor motion as compositional material. Instead of drawing a single static line, it records the timing, geometry, press state, and pressure-like pointer signal of each performance, then arranges multiple takes as a polyphonic score.

Arm the recorder and draw directly on the irregular score field. Each gesture is sampled with bounded frequency, normalized to the field, passed through adaptive smoothing, and resampled at a stable interval. The resulting take retains its own duration, entrance delay, color, path, velocity envelope, and press accents.

Up to twelve independent voices can occupy one score. Selected takes can be mirrored horizontally or vertically, rotated, reversed in time, or quantized against a temporal grid. Canon creates a delayed copy that preserves the original gesture while giving it a new identity and entrance. A multi-track timeline exposes the complete duration and overlap of every voice.

Three visual readings use the same authored data. Paths show the exact sampled gestures and moving playheads. Ribbons translate local velocity into changing width and intensity. Convergence accumulates the number of distinct voices visiting each part of the field. The inspector also computes path distance, duration, point count, peak speed, geometric intersections, and close encounters between simultaneously active cursors.

Optional local sound gives each take an entrance tone and each close encounter a short collision tone. Sound starts only after a user gesture and uses packaged Web Audio oscillators; it does not load audio from the network.

Save Reel stores the score in IndexedDB inside the extension origin. JSON export preserves the editable timing and point data. SVG export produces a self-contained vector plate with embedded animateMotion paths, score measurements, and no external dependencies.

Included capabilities:
• bounded pointer recording with adaptive smoothing and fixed-interval resampling;
• up to twelve delayed movement voices;
• mirror, rotation, reverse, quantize, and canon transformations;
• time-accurate multi-voice playback without rebuilding the DOM per frame;
• velocity ribbons and convergence-density analysis;
• geometric intersection and simultaneous close-encounter detection;
• optional locally synthesized cue sounds;
• IndexedDB save and restore;
• editable JSON and self-contained animated SVG export;
• a complete four-voice bundled score.