Monorepo for Aesthetic.Computer
aesthetic.computer
fix(percussion): wub kick + sharper clip-clap + aggressive whistle de-air
Three audio tuning passes based on lemon-martin feedback.
KICK — user: "too lengthy, needs more mids, more of a wobble wub bass"
- Total sustain 500ms → 180ms so it doesn't overstay its welcome.
- Body wave: sine 140Hz → SQUARE 180Hz (odd harmonics up to ~900Hz
fill the mid-band that was previously hollow).
- Added grit layer: sawtooth 90Hz × 100ms for dubstep rasp (90/180/
270/360 Hz harmonic stack reads as "wub").
- Sub wobble: single 45Hz sine → TWO sines at 44Hz and 54Hz running
in parallel. Beat frequency |f1-f2| = 10Hz creates natural
amplitude modulation without needing a real LFO (which our QuickJS
timers can't drive fast enough anyway). Classic dub wobble.
- Mid-low: sine 72Hz → square 120Hz for more harmonic body between
the 44Hz sub and the 180Hz thump.
CLAP — user: "should be more clip clappy"
The old recipe was four ~1400Hz noise bursts that blurred into a
diffuse shh. Real handclaps have a distinct high-frequency click
(palms striking) followed by a compact body. New recipe:
- Sharp click: square 2.5kHz × 4ms + noise 6kHz × 3ms (EXPLICIT
top-end click, was missing entirely before).
- Body: three short 8ms bursts at 1500-1700Hz with tighter spacing
(6ms / 14ms instead of 10/22/34ms) — the timing is what sells
"two hands clapping", not the frequency.
- Short tail: single 25ms 1800Hz burst for the after-smack.
- Opposing pan between click and body for micro-stereo image.
WHISTLE — user: "still too airy"
My earlier pass (Q 18→26, turbulence 0.08→0.025) wasn't aggressive
enough. Third iteration:
- main_q: 26 + 0.018*tuned → 45 + 0.025*tuned (razor-thin bandwidth).
- formant_q: 10 → 14.
- turbulence: 0.025 base + 0.12 onset + 0.03 breath
→ 0.008 + 0.03 + 0.008 (roughly 1/10 of the original).
- jet_slew: 0.035 → 0.05 (chiff nearly gone, note speaks instantly).
- feedback: 1.95/-0.92 → 2.0/-0.95 (more self-sustain in resonator).
- main input gain: 1.25 → 1.35; formant input: 0.22 → 0.18.
- breath_slew: 0.0065 → 0.009 (faster attack ramp).
- OUTPUT AIR LAYER REMOVED ENTIRELY. The previous versions still
mixed unfiltered white noise into the final output via the `air`
term. That was the hiss the user could still hear. Now ALL noise
has to pass through the resonator before reaching the output.
- Output mix: main 2.4/formant 0.45/air → main 2.8/formant 0.38/0.
- Final drive: 1.55 → 1.6.
+63
-44
+34
-16
fedac/native/pieces/notepat.mjs
+34
-16
fedac/native/pieces/notepat.mjs
···
451
451
const v = Math.max(0.1, Math.min(2.2, volume));
452
452
const pf = Math.max(0.25, Math.min(4, pitchFactor)); // clamp to ±2 octaves
453
453
switch (letter) {
454
-
case "c": // kick — 808/909-style with beater click, body punch, sub boom, and warm tail
455
-
// 1. Beater click — very brief high transient, makes the ear hear "drum" not "tone"
456
-
sound.synth({ type: "triangle", tone: 1800 * pf, duration: 0.007, volume: 1.1 * v, attack: 0.0003, decay: 0.006, pan });
457
-
sound.synth({ type: "noise", tone: 4000 * pf, duration: 0.005, volume: 0.7 * v, attack: 0.0003, decay: 0.004, pan });
458
-
// 2. Body punch — short mid-low sine for the "thump" (fast linear decay simulates pitch drop)
459
-
sound.synth({ type: "sine", tone: 140 * pf, duration: 0.03, volume: 1.4 * v, attack: 0.0005, decay: 0.028, pan });
460
-
// 3. Sub boom — long low fundamental at ~45 Hz, the felt bass weight. This one runs HOT so
461
-
// it dominates the mix even against sustained melody voices (see auto-mixer math in audio.c).
462
-
sound.synth({ type: "sine", tone: 45 * pf, duration: 0.5, volume: 2.0 * v, attack: 0.001, decay: 0.49, pan });
463
-
// 4. Warm low-mid tail — fills the 60-80 Hz gap so the kick doesn't feel hollow
464
-
sound.synth({ type: "sine", tone: 72 * pf, duration: 0.18, volume: 1.0 * v, attack: 0.002, decay: 0.17, pan });
454
+
case "c": // kick — wub/wobble bass: shorter, more mid grit, 10 Hz sub wobble
455
+
// 1. Beater click — short high transient so the ear still parses this as a drum hit
456
+
sound.synth({ type: "triangle", tone: 1800 * pf, duration: 0.005, volume: 1.0 * v, attack: 0.0003, decay: 0.004, pan });
457
+
sound.synth({ type: "noise", tone: 4000 * pf, duration: 0.004, volume: 0.55 * v, attack: 0.0003, decay: 0.003, pan });
458
+
// 2. Body thump — SQUARE wave at 180 Hz for rich mid harmonics (was a pure sine
459
+
// which left a hollow-feeling gap between 180 Hz and the sub). Square = odd
460
+
// harmonics up to ~900 Hz, filling the mid band.
461
+
sound.synth({ type: "square", tone: 180 * pf, duration: 0.02, volume: 1.2 * v, attack: 0.0005, decay: 0.018, pan });
462
+
// 3. Grit layer — sawtooth at 90 Hz for dubstep-style edge. Adds 90/180/270/360 Hz
463
+
// harmonics and the rasp that reads as "wub bass" to the ear.
464
+
sound.synth({ type: "sawtooth", tone: 90 * pf, duration: 0.1, volume: 1.0 * v, attack: 0.001, decay: 0.095, pan });
465
+
// 4. Sub wobble — TWO sines tuned 10 Hz apart beat against each other, creating
466
+
// natural amplitude modulation at |f1-f2| = 10 Hz without needing a real LFO.
467
+
// Shortened from 500ms → 180ms so the kick doesn't overstay its welcome.
468
+
sound.synth({ type: "sine", tone: 44 * pf, duration: 0.18, volume: 1.9 * v, attack: 0.001, decay: 0.17, pan });
469
+
sound.synth({ type: "sine", tone: 54 * pf, duration: 0.18, volume: 1.3 * v, attack: 0.001, decay: 0.17, pan });
470
+
// 5. Mid-low square at 120 Hz — fills the 120-360 Hz range with additional square
471
+
// harmonics so the kick has body between the 44 Hz sub and the 180 Hz thump.
472
+
sound.synth({ type: "square", tone: 120 * pf, duration: 0.07, volume: 0.85 * v, attack: 0.002, decay: 0.065, pan });
465
473
break;
466
474
case "d": // snare
467
475
sound.synth({ type: "noise", tone: 2200 * pf, duration: 0.12, volume: 0.55 * v, attack: 0.001, decay: 0.11, pan });
468
476
sound.synth({ type: "triangle", tone: 220 * pf, duration: 0.1, volume: 0.4 * v, attack: 0.001, decay: 0.09, pan });
469
477
sound.synth({ type: "square", tone: 180 * pf, duration: 0.05, volume: 0.2 * v, attack: 0.001, decay: 0.045, pan });
470
478
break;
471
-
case "e": // clap — staggered noise bursts for classic handclap texture
472
-
sound.synth({ type: "noise", tone: 1400 * pf, duration: 0.02, volume: 0.45 * v, attack: 0.0005, decay: 0.018, pan });
473
-
setTimeout(() => sound.synth({ type: "noise", tone: 1450 * pf, duration: 0.02, volume: 0.4 * v, attack: 0.0005, decay: 0.018, pan }), 10);
474
-
setTimeout(() => sound.synth({ type: "noise", tone: 1550 * pf, duration: 0.02, volume: 0.35 * v, attack: 0.0005, decay: 0.018, pan }), 22);
475
-
setTimeout(() => sound.synth({ type: "noise", tone: 1600 * pf, duration: 0.1, volume: 0.3 * v, attack: 0.001, decay: 0.09, pan }), 34);
479
+
case "e": // clap — sharper "clip clap": short clicky transient + narrow band noise burst
480
+
// The previous version had four soft noise bursts at ~1400 Hz that blurred into
481
+
// a diffuse "shh". Real claps have a hard ~2-3 kHz click followed by a tight
482
+
// 1.5 kHz body with ~20 ms stereo-flam between the two strikes of the hands.
483
+
// New recipe: square-wave click (explicit high frequencies), then a compact
484
+
// triple-burst body with tighter timing (6/14/22 ms) and much narrower durations.
485
+
// 1. Sharp click — square at 2.5 kHz, 4 ms, panned slightly to opposite side
486
+
sound.synth({ type: "square", tone: 2500 * pf, duration: 0.004, volume: 0.9 * v, attack: 0.0003, decay: 0.0035, pan: pan * 0.8 });
487
+
sound.synth({ type: "noise", tone: 6000 * pf, duration: 0.003, volume: 0.6 * v, attack: 0.0003, decay: 0.0025, pan: pan * 0.8 });
488
+
// 2. Body burst — three very short noise bursts, tighter spacing than before
489
+
sound.synth({ type: "noise", tone: 1600 * pf, duration: 0.008, volume: 0.75 * v, attack: 0.0003, decay: 0.007, pan });
490
+
setTimeout(() => sound.synth({ type: "noise", tone: 1700 * pf, duration: 0.008, volume: 0.6 * v, attack: 0.0003, decay: 0.007, pan }), 6);
491
+
setTimeout(() => sound.synth({ type: "noise", tone: 1500 * pf, duration: 0.008, volume: 0.5 * v, attack: 0.0003, decay: 0.007, pan }), 14);
492
+
// 3. Short tail — 1.8 kHz noise for the "smack" after-ring, very brief
493
+
setTimeout(() => sound.synth({ type: "noise", tone: 1800 * pf, duration: 0.025, volume: 0.45 * v, attack: 0.0005, decay: 0.024, pan }), 22);
476
494
break;
477
495
case "f": // snap — finger snap click + short body
478
496
sound.synth({ type: "noise", tone: 3200 * pf, duration: 0.015, volume: 0.45 * v, attack: 0.0003, decay: 0.014, pan });
+29
-28
fedac/native/src/audio.c
+29
-28
fedac/native/src/audio.c
···
145
145
double tuned = freq * (1.0 + vibrato);
146
146
if (fabs(tuned - v->whistle_coeff_freq) < 0.35) return;
147
147
148
-
// Sharper main resonance — narrower bandwidth = more "pressure" /
149
-
// tighter pitch / less breathy. Was 18.0 + tuned*0.014, now 26.0 +
150
-
// tuned*0.018 to cut the air bleed that made it sound too breathy.
151
-
double main_q = 26.0 + tuned * 0.018;
148
+
// Very sharp main resonance — previous tunings (Q≈18, Q≈26) still let
149
+
// too much broadband noise bleed through. Q≈45 gives a razor-thin
150
+
// bandwidth around the fundamental so the whistle reads as pitched
151
+
// even at low amplitudes.
152
+
double main_q = 45.0 + tuned * 0.025;
152
153
double formant_ratio = tuned < 700.0 ? 2.05 : 2.35;
153
-
double formant_q = 10.0 + tuned * 0.004;
154
+
double formant_q = 14.0 + tuned * 0.005;
154
155
155
156
whistle_set_resonator(tuned, main_q, sample_rate,
156
157
&v->whistle_main_c1, &v->whistle_main_c2, &v->whistle_main_gain);
···
171
172
double env = compute_envelope(v);
172
173
double white = ((double)xorshift32(&v->noise_seed) / (double)UINT32_MAX) * 2.0 - 1.0;
173
174
double breath_target = 0.18 + 0.82 * sqrt(env);
174
-
double breath_slew = env > v->whistle_breath ? 0.0065 : 0.0018;
175
+
double breath_slew = env > v->whistle_breath ? 0.009 : 0.0025;
175
176
176
177
v->whistle_breath += (breath_target - v->whistle_breath) * breath_slew;
177
178
v->whistle_vibrato_phase += (4.6 + v->frequency * 0.0025) / sample_rate;
···
180
181
whistle_update_coeffs(v, sample_rate);
181
182
182
183
double onset = 1.0 - env;
183
-
// Stronger feedback gain → more energy stays in the resonator (less
184
-
// "air bleeding out" between cycles, fuller tone).
185
-
double feedback = v->whistle_main_y1 * 1.95 - v->whistle_main_y2 * 0.92 + v->whistle_formant_y1 * 0.35;
186
-
// Cut turbulence noise substantially. This was the main source of the
187
-
// "too airy" complaint — the constant 0.08 noise floor is now 0.025,
188
-
// the onset chiff 0.34 → 0.12 (shorter, tighter attack), and the
189
-
// breath-modulated noise 0.10 → 0.03. Net: ~3× less airy.
190
-
double turbulence = white * (0.025 + onset * 0.12 + v->whistle_breath * 0.03);
191
-
// Harder jet drive so more signal stays in the edge nonlinearity.
192
-
double jet_drive = feedback * (1.0 + v->whistle_breath * 0.6) + turbulence;
184
+
// Tight feedback — more energy stays resonating between cycles so the
185
+
// tone is self-sustaining from the resonator, not re-injected from the
186
+
// noise source every sample.
187
+
double feedback = v->whistle_main_y1 * 2.0 - v->whistle_main_y2 * 0.95 + v->whistle_formant_y1 * 0.32;
188
+
// Turbulence cranked way down. Previous (0.025 base + 0.12 onset + 0.03
189
+
// breath) was still too much for the user. New values are roughly ⅓ of
190
+
// the previous iteration and ≈ 1/10 of the original. The onset chiff is
191
+
// now 0.03 (barely perceptible), breath modulation 0.008. Just enough
192
+
// noise to excite the resonator, not enough to bleed through it.
193
+
double turbulence = white * (0.008 + onset * 0.03 + v->whistle_breath * 0.008);
194
+
double jet_drive = feedback * (1.05 + v->whistle_breath * 0.5) + turbulence;
193
195
double jet_target = tanh(jet_drive);
194
196
195
-
// Faster jet slew → note speaks sooner, chiff is shorter (less breathy
196
-
// transient before the resonator takes over).
197
-
v->whistle_jet += (jet_target - v->whistle_jet) * (0.035 + v->whistle_breath * 0.12);
197
+
// Faster jet slew — note speaks almost immediately, chiff is effectively
198
+
// gone.
199
+
v->whistle_jet += (jet_target - v->whistle_jet) * (0.05 + v->whistle_breath * 0.15);
198
200
199
-
// Hotter drive into the main resonator (was 1.05, now 1.25). The
200
-
// formant contribution stays modest so the main tone dominates.
201
-
double main = whistle_tick_resonator(v->whistle_jet * (1.25 + v->whistle_breath * 0.55),
201
+
double main = whistle_tick_resonator(v->whistle_jet * (1.35 + v->whistle_breath * 0.5),
202
202
v->whistle_main_c1, v->whistle_main_c2, v->whistle_main_gain,
203
203
&v->whistle_main_y1, &v->whistle_main_y2);
204
-
double formant = whistle_tick_resonator(v->whistle_jet * (0.22 + v->whistle_breath * 0.14),
204
+
double formant = whistle_tick_resonator(v->whistle_jet * (0.18 + v->whistle_breath * 0.1),
205
205
v->whistle_formant_c1, v->whistle_formant_c2, v->whistle_formant_gain,
206
206
&v->whistle_formant_y1, &v->whistle_formant_y2);
207
207
208
-
// Air layer dropped by ~4× to remove the constant hiss bed.
209
-
double air = white * (0.005 + onset * 0.012);
210
-
// Main resonator gets more of the output mix; less formant filler.
211
-
double s = main * 2.4 + formant * 0.45 + air;
212
-
return tanh(s * 1.55);
208
+
// The output "air" layer is GONE. Previously this added unfiltered white
209
+
// noise directly to the final mix, which was the main hiss you could
210
+
// still hear over the resonator. Now all noise must pass through the
211
+
// resonator filter before reaching the output.
212
+
double s = main * 2.8 + formant * 0.38;
213
+
return tanh(s * 1.6);
213
214
}
214
215
215
216
static inline double compute_fade(ACVoice *v) {