Hot-swap explained
Hot-swap is what turns NKIDO from a synth engine into a live-coding instrument. You change the code, press a key, and the new patch takes over while keeping oscillator phase, filter state, and delay lines intact.
The problem
A naïve “stop the old patch, start the new patch” approach sounds terrible:
- Oscillator phase resets → audible click.
- Delay lines and reverb tails drop to zero → tail cuts off mid-reverb.
- Envelopes restart → note retriggers unexpectedly.
A live-coding system needs to diff the new patch against the old one, keep the nodes that still exist, and smoothly fade in anything new.
Semantic IDs
Every node in a NKIDO patch gets a stable semantic ID derived from its position in the source plus its operator + constant args. If you change a filter’s cutoff but leave everything else alone, the node’s ID stays the same. NKIDO recognizes it and keeps its internal state.
What gets preserved
When a node’s ID matches between old and new:
- Oscillator phase.
- Filter delay registers (biquad state).
- Delay-line buffers and read/write positions.
- Envelope stage and level.
- Stateful RNG seeds for noise generators.
New nodes start fresh. Removed nodes are phased out with a short crossfade.
When IDs change
Some edits necessarily break the identity of a node: adding a new filter in the middle of the chain, changing an oscillator’s waveform, or renaming a variable. In those cases NKIDO falls back to a short crossfade (~10ms) to avoid clicks.
This is why live-coding feels musical in NKIDO: you can sculpt a running sound by nudging parameters, and the changes land on the beat instead of restarting the whole patch.