autosynth

This trained object

no world loaded

Receipt

armed—

disarmed—

σ_φ (cont/dens/nov)—

Raw audio never leaves the device (CS-1). Only stage-3 anchors cross the wire; σ_φ is measured locally and embedded in the world.

Source Audio

Drop audio here or click to choose

Local run: files stay on this machine. Hosted run: files upload to this app's server for training (see the privacy note).

no files ingested yet

Local sealed mode: only gauge-invariant stage-3 cost matrices leave this box. Hosted mode: your audio is processed on this app's server (operator's R3 model).

Train (cloud anchor-fit)

ingest at least one source, then train.

ingest

stage-3 encode

cloud anchor-fit

verify receipt

build index

σ_φ calibrate

save world

Share this set

off — your trained set is private. Toggle to list it in Explore.

Instrument world not loaded

Connecting to the companion…

Open a shared set from Explore to listen and steer, or ingest audio + Train on cloud to build, play, and steer your own trained world right here. Training also measures a σ_φ calibration for your corpus, so after it finishes the instrument switches to your world — live to play and steer.

steering surface no world loaded

no world loaded

your force achieved

earned words

Knobs extra modes · TEMP throttle · GAUGE frame

Lane Console display · live arrangement statistics (read-only)

settle0.0

gauge[g]

slide—

loop—

output0.0

legacy lanes (VARY / SPREAD / DENSITY)

drag to lean · handle = your target, mark = what the engine achieved

Source Library 0 loaded

no sources ingested

Output Tape playhead from telemetry · waveform not wired

00:00 tape · live window --:--

Explore — shared sets

Open a shared set into Play and steer it with the same region-tilt lane. One instrument, many sources.