Step 01 / 5
Build a lightly spread saw lead #
Set giStage init 1 in the full script. Mix three band-limited saws at −5, 0 and +5 cents, each at −9 dB. Their initial phases are 0.11, 0.39 and 0.72 cycles. Weight the voices 0.8/0.5/0.2 on the left and reverse those weights on the right; divide each channel sum by 1.5.
Use a linear amplitude envelope with 4.40 ms attack, 643 ms decay, −4 dB sustain and 191 ms release. Instrument gain is −6 dB, and touch scales amplitude by 0.4 + 0.6 × velocity. This contour gives a quick lead onset and a short fade without making every note a click.
portk smooths pitch toward each new target with a 40 ms half-time, starting from the preceding note's pitch. The score uses sequential notes; this simple pitch memory does not implement Live's mono-legato priority. The saws interpret Basic Shapes at position 66%, rather than scan its actual wavetable.
Sound after step 1 — WebM/Opus
Next: Add the upper color, sub and moving notch →Step 02 / 5
Add the upper color, sub and moving notch #
Set giStage init 2. Each unison voice gains a triangle at −13 dB and +5 cents. This triangle is an explicit interpretation of the second Basic Shapes position. Add a centered sine sub at −18 dB, one octave below the note.
kFilter linsegr 0, 0.020, 1, 1.10, 0.20, 0.300, 0
kNotch = limit(1090*pow(2, -0.8*kFilter), 30, 18000)
aLeft butterbr aLeft, kNotch, kNotch/1.2
Apply the same notch on the right, then lowpass each channel at 6800 × 2^(0.35 × velocity) Hz, limited to 18000 Hz. The negative 0.8-octave envelope sweep hollows the lead during its early body; the lowpass limits the upper edge. Its velocity route brightens stronger notes. These octave depths and notch bandwidth interpret the source matrix and resonance values. Mod-wheel and pitch-bend routes are not mapped.
Sound after step 2 — WebM/Opus
Next: Thicken the lead with smooth clipping →Step 03 / 5
Thicken the lead with smooth clipping #
Set giStage init 3. Remove DC before the shared nonlinear stage. Use tanh(input × 3 × ampdb(3.1)) / 3 at 70% wet, blended with 30% clean signal, then trim the result by −3 dB.
The +3.1 dB drive pushes louder peaks farther into the rounded part of the curve. The factor 3 sets its sensitivity; dividing by 3 preserves the curve's small-signal scale before drive. This is a smooth interpretation of Analog Clip, not its exact transfer curve. Processing the shared note bus lets overlapping releases affect the drive together.
Sound after step 3 — WebM/Opus
Next: Add filtered ping-pong repeats →Step 04 / 5
Add filtered ping-pong repeats #
Set giStage init 4. Send the mean of the two dry channels into a 267 ms left delay. Feed 24% of each filtered echo into the opposite delay; the right side first receives the left echo rather than a second copy of the original input.
aEchoL vdelay3 aSend+aFeedbackR*0.24, 267, 1000
aEchoR vdelay3 aFeedbackL*0.24, 267, 1000
Highpass and lowpass each echo around an 1800 Hz center, at center / 2^1.75 and center × 2^1.75. This interprets Width 3.50 as a 3.5-octave passband. Mix 82% direct and 18% wet. The first echo is left, and subsequent feedback traversals alternate sides. Each feedback traversal also includes one 16-sample control block. No delay-time or filter LFO is active.
Sound after step 4 — WebM/Opus
Next: Balance the finished stereo lead →Step 05 / 5
Balance the finished stereo lead #
Set giStage init 5. Apply +5.80 dB to each output channel and remove DC using dcblock2. Keep the existing stereo difference; there is no bass narrowing or mono conversion in this version.
The gain raises both the lead and its repeats. Clear the shared input buses after processing so previous note samples do not accumulate. The finished sound has a fast, hollowed lead body with a centered sub and short, alternating echoes.
Sound after step 5 — WebM/Opus
Audio & settings ↓