Step 01 / 5
Build the drifting triangle cluster #
Set giStage init 1 in the full script. Use five triangle voices spread across ±24 cents, with independent sine drift of ±2.8 cents. Their different phases and slow rates make the held spectrum beat. Use engine gain −10 dB, amplitude attack 0.65 s, decay 3.2 s to −4 dB and release 1.9 s. A −8 dB instrument gain and 50% velocity response follow mild smooth drive. Spacing and motion are explicit Csound choices for the source macros.
| Csound control | Value |
|---|---|
| Triangle offsets | −24, −12, 0, +12, +24 cents |
| Drift | ±2.8 cents; 0.12 + abs(offset) × 0.005 Hz |
| A contour | 0.65 / 3.2 / −4 dB / 1.9 s |
| Engine / instrument gain | −10 / −8 dB |
Sound after step 1 — WebM/Opus
Next: Add the slower octave and stereo filtering →Step 02 / 5
Add the slower octave and stereo filtering #
Set giStage init 2 in the full script. Enable the octave-up sine at −22 dB and +3.125 cents. Give it an independent 1.10 s attack, 4 s decay to −6 dB and 2.20 s release. Filter the triangle channels at 2600 Hz with Q 1.2; reverse the voice weights across the channels, bias A slightly left and B right. The octave enters later than the main cluster and can remain after its release. This interprets Basic Shapes and SVF rather than scanning the original engines.
| Csound control | Value |
|---|---|
| A lowpass | 2600 Hz, Q 1.2 |
| B pitch / level | 2 × note frequency, +3.125 cents / −22 dB |
| B contour | 1.10 / 4.00 / −6 dB / 2.20 s |
| Stereo voice weights | 0.8/0.7/0.5/0.3/0.2, reversed right; sum / 2.5 |
Sound after step 2 — WebM/Opus
Next: Add filtered stereo echoes →Step 03 / 5
Add filtered stereo echoes #
Set giStage init 3 in the full script. Use 340 ms left and 470 ms right, 30% feedback and 22% wet. Inject the centered input on the left and cross-feed the filtered echoes so later repeats alternate sides. Filter each echo around 1700 Hz, with a 3-octave interval between highpass and lowpass cutoffs. A shared 0.18 Hz sine moves times by ±2% and the filter center by ±0.04 octaves. Feedback values are updated once per 16-sample block, adding that small delay to the recirculating path. Held notes leave older colors behind the current tone.
| Csound control | Value |
|---|---|
| Times | 340 / 470 ms |
| Feedback / wet | 30% / 22% |
| Routing | Crossed, left input |
| Filter center / width | 1700 Hz / 3 octaves |
| Time LFO | 0.18 Hz, ±2% |
Sound after step 3 — WebM/Opus
Next: Surround the tone with a damped room →Step 04 / 5
Surround the tone with a damped room #
Set giStage init 4 in the full script. Feed the preceding signal through 30 ms predelay into reverbsc, with feedback 0.8089, 3200 Hz damping and 20% wet mix. Predelay leaves the initial articulation in front; feedback prolongs the tail, while damping darkens repeated circulation. The source room decay setting is 3.3 seconds, but Csound feedback is not a decay-time control: this is an interpretation of that space, with different diffusion and modulation.
| Csound control | Value |
|---|---|
| Source decay target | 3.3 s (not a measured Csound decay) |
| Predelay / feedback | 30 ms / 0.8089 |
| Damping / wet | 3200 Hz / 20% |
| Room side multiplier | 1 |
| Room input cuts | Off |
Sound after step 4 — WebM/Opus
Next: Set the final width and gain →Step 05 / 5
Set the final width and gain #
Set giStage init 5 in the full script. Use 16.5 dB gain after the effects. Scale the side signal by 1, preserving the center. Highpass only the side at 160 Hz to focus the lowest frequencies. Apply order-2048 DC removal, using a longer window to limit low-note loss. The fixed output-scale control at the end of instrument 90 remains a separate Csound gain setting. Clear the shared note buses only after processing them; effect memories continue carrying their own tails.
| Csound control | Value |
|---|---|
| Final gain | 16.5 dB |
| Width | 1 × side |
| Bass side cutoff | 160 Hz |
| DC removal | Order 2048 |
Sound after step 5 — WebM/Opus
Audio & settings ↓