Spectral Resonator: Make a Pitched Texture from Noise

Draw out the space between bursts

Play the example and concentrate on the gaps. The source is an original eight-second synthesized noise pattern, with a low noise floor between louder bursts. Its broad spectrum offers energy across many analysis bands. The effect organizes that energy around Freq at 220 Hz and a set of harmonics, then lets the resonant energy decay.

Move Dry/Wet to zero briefly to identify the source, then return to 85%. The contrast to follow is broadband rushing noise versus a more pitched, lingering texture. This is not evidence of a hidden synth playing extra notes. Audio is the excitation, and the implementation offers internal frequency control only: there is no MIDI sidechain or external note input for this device.

Three different ways to change the cloud

Freq: where the harmonic pattern starts

Compare 220 Hz with 440 Hz while leaving the other settings alone. The fundamental target moves up an octave. This does not transpose every feature of the original noise; it moves the resonant pattern that selects and sustains spectral energy. Return to 220 Hz before the next comparison.

Harmonics: how much upper structure is available

Reduce 12 to 3, then restore it. Fewer partials make a simpler tonal shape; more can supply a richer upper spectrum. Turning the count up is not a substitute for useful source energy. Stop adding harmonics when the upper ring masks other bright parts or distracts from the lower pitch.

Res: how sharply the pattern is emphasized

Compare 70% with 25%. Higher settings narrow the harmonic emphasis; lower settings allow a broader, less exclusively tonal texture. Listen for definition versus diffuseness, not just a level increase. A sharply isolated whistle is not necessarily better than a softer halo.

Choose the time scale before the final level

Decay governs persistence. Try 35% against the initial 65% and listen to whether one burst still hangs over the next. Long settings make rhythmic input act like a sustained layer. Shorter settings preserve more of its separate gestures. The percentage is a control amount, not a duration in seconds, and the mapping is nonlinear; halfway is not half of the longest possible ring time.

After setting Decay, revisit Gain and Dry/Wet. Greater persistence increases the amount of sound occupying a gap, which can be mistaken for an improvement in tone. Compare at a similar perceived level. If the texture feels impressive alone but clouds a rhythm, reducing the wet share may help more than making the resonance still sharper.

Why the tuning is not infinitely precise

FFT selects 1024, 2048, 4096 or 8192 analysis samples. Larger sizes provide a finer frequency grid at the cost of more latency. The pitched output is quantized to that grid, so Freq is a requested fundamental rather than a guarantee of exact oscillator tuning at every setting. Very small frequency changes may not behave like a smooth synthesizer glide.

Changing FFT rebuilds the analysis engine and clears its resonance state. Stop playback, choose another size, and compare a fresh burst instead of sweeping the control through a sustained tail. Keep 4096 for the main experiment so a structural reset is not confused with a Decay response. If the musical task needs a precisely specified chord of independent voices, this fixed harmonic pattern is the wrong starting point; use a device with explicit per-voice tuning.