AURORA
ideas
What is asked for, and where it stands
Players ask for things, and some of them are better than what was planned. This page keeps the list in the open: what is next, what is planned, what is being thought about, and what was decided against and why. To add to it, open an issue.
next being built now planned agreed, in the queue (nothing is marked next until the next one is picked) thinking wanted, shape undecided not this decided against
The sound
A rhythmic gate on the whole sound, synced to the MIDI clock like the delay: depth, a note division, the edge from soft to hard, a few patterns and one that drifts. Before the delay and the plate, so the tail fills the gaps. A fourth FX page.
asked for by a player on Reddit
A wavefolder on the METAL page: where DRIVE flattens the peaks, this bends them back, growing inharmonic overtones the harder you play. The scream under a horn.
The air through a resonator tuned to the note: breath through a pipe, a bow on a string, a struck drone. The other empty knob on METAL.
An anvil, a scraped drone, a riser, once FOLD and RESO are in.
Scenes and sounds
Each sound would carry its own B, so a pair travels together, and you could still point B at any other preset or USER slot, A or B side. Today B is one setting for the whole instrument, picked with ARP and PRESETS. The question is what the PRESETS encoder shows when every slot has two halves.
asked for by a player
The morph with more stops: a chain of scenes, each held for a time and crossfaded into the next, so sounds come in and go out on their own over a held chord. AURORA stays a sound module (your sequencer plays the notes); this would sequence the sound, not the notes. The step clock could follow the MIDI clock like the delay. The open question is how it is edited on four knobs.
asked for by grufallokoster on Reddit
FREEZE's grains played backwards, and a reverse on the delay: the sound sequenced in reverse, swelling up to where it was struck. Cheap, since the grains already read from the delay line.
asked for by grufallokoster on Reddit
SAVE then GLO sends the scene you are at. A pair dump would send A and B together, marked, so a two-layer sound travels in one message and one issue.
The share page sending a sound to the FM-1, and the emulator taking the text Copy sound makes, so sounds go both ways and can be heard before they are installed.
The instrument
Some settings together click on the FM-1: most likely a block of sound taking longer than its 5.8 ms when eight voices, METAL, DRIVE, the reverb, the delay and FREEZE are all busy at once. The MIDI page (GLO) shows LOAD for both cores; the fix is a guard that lightens what costs most when a block runs over, instead of letting it click, plus a leaner worst case. If you hear it: which preset, which knobs, and the LOAD reading would settle where the cost is.
spanasik on Reddit
Today the output is a soft knee at about -4 dBFS, a chord gain that eases eight notes down by up to 7 dB over half a second, and DRIVE, which compresses on its way to grit. A big chord into a long reverb can lean on that knee and breathe. A BUS page under FX would put a low and a high shelf and a real limiter, with its own attack and release, after everything; the cost is a block of CPU that the heaviest settings do not have yet, so the clicks come first.
spanasik on Reddit
The engine is plain C and already runs on a computer and in the browser. A VST3, AU and standalone wrapper would run it at 44.1 kHz and resample, with every control automatable and the screen as the GUI. Signing for Macs is the awkward part.
The FM-1's radio can do BLE MIDI with an open stack at a modest cost in flash. One link, notes and CCs, no audio. Worth it for a phone or a tablet as the controller.
The battery's voltage is on a pin nothing reads. A mark on the MIDI page.
Needs the closed Wi-Fi libraries, most of the remaining flash and memory, and giving up USB audio. Not for this firmware.
The pad is built from sine overtones; TILT, SPREAD and BRIGHT already are the shape, and METAL, DRIVE and the swarm add what a saw or square would. A waveform switch would add nothing the knobs do not.
Recently done
A click left half-finished used to teach the encoder a second detent, and every click after that counted twice. The power-on position is the detent now, and a second one is only believed when the knob keeps resting there.
veecheech, and the author
A VOICE page under ENV: how many notes sound at once, 1 to 8, saved with the sound. At 1 the last note wins, GLIDE is legato and a key still down comes back. The horns and PULSAR ship at 1.
veecheech
USER slots are read by the length of the sound when they were written, so a new control loads at its default instead of wiping the bank. Slots from 0.7.10 carry over.
A click on its own moves one, quick clicks move two, so every value is reachable; detents were learned slowly, which was not enough (see 0.7.13).
veecheech
And B's edits survive a power-off.
veecheech
The first two sounds sent in through the share page, now presets 24 and 25.
veecheech
SAVE then GLO sends the sound over USB MIDI; the share page turns it into text.
The keys' own touch, poly aftertouch out, and the delay following the clock.