INTERVENTION / MEMORY / AUTONOMY CONTINUES
DWELL 00:00:00 · AGE 00:00:00
SEED 0000000000 · RECALL NONE / 0
NO MUTATION
SYSTEM RETAINS ITS CURRENT CONDITION
PITCH / ERROR / RESPONSE / EVENTS / RESTS
32×16 LOGICAL FIELD / SOUND · MEMORY · NETWORK · HUMAN
SOUND SCULPTING THE INTERNET
EVENTS / RESTS · 0 / 0 VISIBLE
0 ACTIVE · 0 EVOLVING · DIST 0.000
SOUND SCULPTING THE INTERNET / TEST.049 / SET.00001
Use audio you own or may transform. Public HTTPS links must allow browser access. Files stay in this browser. Maximum 12 MB; up to 12 seconds retained. Codec support varies by browser.
Source discovery begins before audio output is unlocked. Each accepted excerpt becomes a buffer, wavetable, resonator profile and evolving response body.
No audio analysed
AWAITING SOURCE ANALYSIS
SCAN HEADS / -- · -- · -- · --
SOURCE DATA / awaiting material
SOUND explores the source and spectral stretch.
CONTROL changes scan novelty, resonance and memory.
All four forms remain available simultaneously.
NO ACTIVE RECALL
X: SOURCE + VOWEL + TRANSIENT · Y: STRETCH + DECAY
X: NOVELTY + BRANCHING · Y: REPETITION + RESONANCE
Evolving convolution: the impulse response is the instrument body.
Three stereo taps return through a filtered, bounded feedback path.
The system is already active. Your first click, touch, Enter or Space becomes a HUMAN encounter and unlocks its audio output. Move SOUND to disturb pitch and the vocal body; move CONTROL to perturb stability and repetition. Release a gesture and let the system return to it.
Hear excitation temporarily bypasses the bodies for comparison. Space: pause / resume · F: freeze · Esc: silence.
Discovered sound → buffer scan + wavetable + resonators → convolution body.
Four scan heads traverse the same excerpt at independent speeds. Its spectral frames become a wavetable and a diverse filter bank, while moving fragments form the source-linked impulse response.
neuro/SPEC is an artistic behavioural system informed by attention, sensitivity, association and repetition, not a clinical model. Rising load reduces excitation; absence shifts attention towards remembered gestures. The response uses generated contours mixed with windowed captures of the synth itself, not measured room acoustics.
One accepted internet source is analysed into eight spectral frames, four independently moving buffer heads, twelve source-derived resonant regions and a moving impulse response. These forms remain related to the source but can be mixed separately. Its measured 64-point energy and roughness contours create two linked NET/SPEC rhythms. The slow macro grammar moves through ACTIVE, REST and ACCENT; within audible macro periods, a repeatable 16- or 32-step micro pattern adds faster openings and rests. Energy drives presence strongly, while roughness can raise subdivision density and shorten openings. Both layers continue traversing the source, and all source analysis and synthesis continue underneath audible rests. New responses crossfade into the old ones. The main synth combines independently mixed oscillator, filtered-air, vocal-tract and internet-wave layers. Its six palettes alter oscillator ratios, portamento, envelope duration, vocal excitation and the spectral weighting of the source-derived wavetable without stopping the sequence. A separate monophonic cell layer repeats defined atonal contours in odd phrase lengths, preserving anchor notes while network data introduces bounded deviations between them.
Error material adds four long polymetric lanes. FAULT, CRC, ALIAS and DROPOUT use held and quantised noise, micro-buffer discontinuities, packet repeats, event loss and source-derived resonances. Their events open four parallel materials: synthetic transients, web-derived shards, moving vowel ghosts and resonator/convolution tails. The sound sets retune and rebalance those materials while the rhythm topology continues. The autonomous composition can occasionally let one unlocked lane borrow an adjacent sound set while the other three retain the current base kit; the variation can later return to the base. CRC’s two bright noise bands carry a slow bounded filter drift across otherwise unchanged events. Error rate introduces repeat bursts and missing events; random walk, random jump, source data and joystick motion alter timing and topology without abandoning the editable pattern.
The internal motion model sends its measured motion back into the response generator and patch matrix. Freeze behaviour holds internal motion and adaptive correction; Hold imprint locks the convolution responses. Four polymetric lanes supply notes and excitation. Four further lanes shape the effect environments; their BODY event target reconstructs the response rather than changing a conventional insert effect. SOUND and CONTROL automatically retain meaningful human gestures in a finite local memory. Each trace weakens and disappears over approximately three minutes; recall can reintroduce its influence without making it permanent. TEST.019 adds a slow seeded compositional drift above these existing systems. It changes one bounded relationship at a time, waits long enough for that state to be heard, respects lane locks, and may partially recall an earlier rhythm, timbre relation or NET interpretation. Rhythm structures enter only at their own cycle boundaries; timbre and NET interpretation move gradually around the preserved TEST.018 centre.
This is an original experiment informed by Synthi 100 routing, Neuron timbral exploration, Notron sequencing, crusher-X feedback and nonlinear generative percussion. It is not a hardware emulation, neural model or claim of a newly invented synthesis algorithm. NET/SPEC searches Wikimedia Commons across more than 190 rotating search terms and accepts supported, permitted audio into a bounded session cache. Source links and credits remain visible. Optional DAW MIDI and additional source adapters are later stages. TEST.020 adds a deterministic 32×16 LIGHT interpretation beside sound. Its four upper rows compress the actual error-lane patterns; the remaining cells interpret SELF, MEMORY, live acquisition, source-derived NET macro/micro rhythm, compositional drift and bounded HUMAN influence. TEST.024 adds persistent spatial symmetry that HUMAN and fading MEMORY activity can break locally, and separates the luminous browser preview from a conservative future hardware adapter scale. TEST.025 retains the modulation engine in development view while the public routing territory becomes a dense read-only cross-talk field made from actual system state; it also raises the existing bass root by two semitones and adds slow bounded CRC filter motion. TEST.027 reverses that bass change by restoring the complete TEST.015 bass engine and its 41.2 Hz root, while isolating its target order and timing from later autonomous composition until listening confirmation. TEST.028 keeps that engine byte-identical and adds only a separate short phrase-breath layer after it: large transitions in the established target sequence briefly withdraw the bass while all three portamentos continue moving underneath. TEST.029 adds a separate irregular eleven-step phrase lane of full, veiled and silent states; unequal step lengths drift against the twelve-note target order while leaving its pitches and portamentos untouched. TEST.030 compacts the existing screen territories, gives LIGHT and NET/SPEC equal width, joins cross-talk and Telex into one lower band, and exposes that eleven-step bass phrase as read-only screen telemetry. TEST.034 changes only the first human audio-unlock gesture from pointer-down to trusted click/tap activation for mobile browser compatibility; synthesis and composition behaviour are unchanged. Its logical row-major output is dimension-independent and separate from any physical panel, wiring or transport implementation. The system restores memory, evolves behaviour and may search and analyse material before its audio output is unlocked; the first human encounter unlocks sound. Gesture history remains in this browser only and is not shared between visitors. Audio pauses when this tab is hidden. System age measures local history, not uninterrupted playing time.
Technical references: Julius O. Smith — formant synthesis · Web Audio convolution