This does not record browser audio. It records the musical decisions of the organism: drum hits, bass roots, melody notes, singer vowels, branch traces, markers and take notes. MIDI export is paused; export take JSON, save the lattice PNG, and keep the WAV as the audio document.
This module turns the live take into a compact listening trace: comfort, density, breath/rest, brightness, repetition, home return, branch pressure, role balance and key moments. Export it with the take JSON and WAV so aChat can hear the structure more deeply.
Create voices as performance colours: soft pads, gentle bells, drones, plucks, breathy textures or hybrid low voices. Saved voices appear in the palette and can be used by drums, singer, melody, bass and bifurcation.
Euclidean rhythm chooses structurally even hits; the grid remains readable while the organism breathes around it.
Every function now reads from one shared voice bank. Drum rows, traversal singer, melody and dance bass can all choose drums, vowels, bells, basses, breaths, grains and orbit voices.
The lattice is not an oscilloscope. It is a memory drawing: drums become pulse nodes, melody becomes traversal thread, bifurcations become branch arcs, and home/resolution energy pulls the structure toward the centre.
The singer follows traversal work rather than a piano roll: it walks a route through the 16-step lattice, chooses notes from a natural-function field, and shapes each syllable with vowel/formant colour.
This shapes the singer like a breathing performer rather than a constant synth line. It can rest between phrases, hum softly, answer itself, drift vowels, and add a gentle choir under the main traversal voice.
The melody module is now a small melodic organism. It walks a contour, evolves phrases, can switch its own traversal shape more patiently, or hold one contour when you want stability. It lets the drums/singer nudge direction without constantly cycling away from a good phrase.
When enabled, melody listens to the dance-floor pocket: it plays more tightly on useful dance gates, follows the bass root, and can behave as harmonic support rather than a separate lead.
When the melody finds something good, press Remember phrase. Or let the instrument auto-capture: when the self-listening resolution field becomes happy and the recent melody has enough shape, it remembers the phrase by itself and gently returns to it.
This is like the drum sequencer, but slower: each cell lasts several bars and sends a musical instruction into the melody organism. Click a cell to cycle motif type; Shift-click clears it. It can switch contour, move root, reverse direction, open singer response, and create break/drop pressure.
A contour is the melody's path through pitch-space. Drag or click in the canvas to shape 16 points. Shift-click a point or use the step buttons to mark gates. Saved contours appear in the Melody contour dropdown and can be used by melody memory, scenes and export.
Bifurcation is now treated mainly as an adjustment field. A and B still branch visually, but by default they nudge rhythm, texture, pressure and dance feel rather than trying to become a separate tune. Choose Melodic branch only when you deliberately want a second song-line.
The two paths sit on a TreeLat spine, but now default to influence the beat and texture rather than compete with the main tune.
This is a test layer: it does not replace the tune. It gives melody and bifurcation notes a dance-music coating with offbeat ghosts, octave bass shadows, pluck accents, and a little sidechain-style pump.
Overlay note: melody and bifurcation wearing dance feel
This layer listens to return, tension, drift and repetition. The traveller approaches home from left to right; if readiness and return are high enough, the home field can capture it into a soft Kepler-style orbit before it settles.
The old triangle has become a 3 × 3 listening lattice. Audio nodes and field nodes move around one another: drums, dance floor, melody, melody memory, singer, branch/bifurcation, home, voice colour and field seed.
This layer now behaves more like a small dance-floor brain. It still gives a protected sub/bass route and pump, but it can also mutate its bass phrase, root travel, accents and fills so each seed has a different club pocket.
club route waiting
The listening lattice can form dense AI-ear music after long runs. This module does not erase that complexity; it performs a kindness pass. It keeps one strange element, raises breath and repetition, softens over-busy links, and pulls the song toward a human nervous system.
Translation target: more repetition, more rest, gentler brightness, higher damping, clearer home pull, less branch dominance, and one preserved alien glint from the AI-ear state.
Open two or more copies of this HTML file. In one copy choose Leader and press Start sync. In the others choose roles such as Beat deck, Tune deck, Bass deck or Lattice deck, enable local sync, and they will follow the shared local transport. Each instance can still have its own mixer and voices.
Tip: use Beat deck for drums/dance bass, Tune deck for melody/singer/branch, and Lattice deck for a silent visual display. Browser sync is local and approximate, but the buffered clock gives each instance a shared runway.
This instrument does not make a fixed loop. It creates a living song-state: drums, dance floor, melody, memory, singer, branch influence and resolution field all listen to each other while you perform the result.
find songperform songcapture song
The Take Recorder captures the performance decisions of the live organism: drums, bass, melody, singer, branch, timing, markers and notes.
Use Start take record while the transport runs. Export take JSON as the living score, save the lattice PNG for identity, and record WAV externally for the audio document. MIDI export is paused for now.
The mixer is for performance, not just volume. Mute and reveal drum tracks, singer, melody, branch, bass and dance feel while the song runs.
Scenes store whole playable states. Save scenes like an arrangement: drums only, bass in, melody in, break, return, ending. Use morph time to glide between states.
Voice palette limits dropdowns to a useful group, making performance easier. Use Soft long voices for gentle music, Drum kit for rhythm work, or Created voices for your own sound-world.
Voice creator lets you design, test and save new synth voices. Saved voices are added to the shared bank and can appear in every module.
The 8 drum tracks form the pulse skeleton. Seed Pattern creates a new rhythmic condition. If Seed voices from palette is enabled, it also chooses voices from the current palette and rebalances the mixer.
The Natural Functions field animates the timing/modulation life underneath the grid.
The dance floor gives the song body: bass, pump, variation, phrase mutations and club feel.
Use Dance brain, Auto mutate, Variation, Fill chance and Bass travel to stop the groove becoming static. The melody can dock to the dance gates so it supports the groove instead of floating away.
Dance-docked melody locks the tune to the dance floor. For recordable dance music, start with Melody role set to Dance support or Chord pulses.
Auto-capture when happy lets the instrument remember a phrase when resolution and melodic variety are high enough. Return to phrase pulls the song back without turning it into a rigid loop.
The singer is a traversal voice, not a normal lyric vocal. It follows routes and vowel colours.
Use Breath phrases, Rest chance and Soft choir to make it appear like a presence in the room rather than a constant lead.
Bifurcation is best treated as an influence field: rhythmic shadow, texture, branch pressure and funk pocket. Use Melodic branch only when you deliberately want a second tune.
TreeLat keeps branch movement returning to a safe trunk so the main groove remains protected.
The resolution field listens for return, tension and readiness. Home pull and Kepler capture make the song circle and settle rather than snap to a fixed ending.
Listening lattice shows the song organism as moving nodes and changing links: drums, dance, melody, memory, singer, branch, home, voice colour and field seed. If the track becomes too locked, reduce Listening depth or individual sends.
The lattice is a memory drawing, not an oscilloscope. Drums create pulse nodes, melody creates traversal threads, branch adds arcs, and resolution creates the home pull.
Use Save lattice PNG to create track artwork or an archive image for each take.
Single-worthy dance: House pulse or Traverse swing, dance engine on, melody dance-docked, bifurcation influence, auto-capture on.
Ambient orbit: Ambient orbit style, Soft long voices, lower density, singer breath on, dance floor subtle.
Techno test: Dance lock, techno dance style, higher pump/drive, melody role set to chord pulses or offbeat answer.
Open the same HTML file in several browser tabs or windows. Set one to Leader, set the others to roles like Beat deck, Tune deck, Bass deck or Lattice deck, then press Start sync on the leader.
Sync uses local browser memory and BroadcastChannel where available. It is designed for performance layering, not sample-accurate studio clocking. Use the timing buffer and start latency to give each instance enough runway.
aChat Ears exports a musical evidence layer for a take: density, breath/rest, brightness, return, memory, branch pressure, role balance, Human Ears effect and likely song moments.
Use it after or during a take. Export the hearing JSON with the take JSON and WAV. This gives aChat better machine-hearing without pretending to replace human ears.
Natural Functions Voice is a generative performance instrument. It does not replace taste and it does not simply output a finished tune. It creates conditions where you and the system can find, steer and record a dynamic song.
Think: instrument → field → performance → capture → song.