An edit is a decision list
The editor arranges two original recordings into clips with source bounds, timeline positions, gain, fades, and speed. Original bytes remain immutable. D3 scales and SVG shapes place the ruler, waveforms, and envelope controls.
Dragging, numeric controls, split, trim, crossfade, mute, solo, and undo all operate on the same session model. The waveform is explicitly described as a navigation aid rather than a calibrated loudness meter.
Make rendering predictable
The server trims at source sample boundaries, resets timestamps, resamples into a chosen mix rate, applies speed stages, fades, gain, and delay, then mixes. A latency-compensated limiter is optional. The documented export formats include WAV, FLAC, and MP3 with explicit settings.
The final endpoint can be set independently of the natural clip duration. Short ramps at the beginning and end reduce abrupt boundary pops.
A built-in demo synthesizes two short musical recordings in the browser and imports them through the same server pipeline used for other audio.
Respect environment and ownership
Local mode explains when a renderer is missing; hosted mode refuses to start without the required storage capability and renderer. Session saves detect concurrent edits, and original downloads are ownership checked. The audio pipeline and its editing model therefore share a clear boundary around what is available, what belongs to an account, and what has actually been saved.