Home›Articles›Felucca MPE: per-note expression firmware for the M-VAV…
Articles

Felucca MPE: per-note expression firmware for the M-VAVE FM-1

Felucca MPE adds per-note pitch bend, pressure and slide to the M-VAVE FM-1. What it does, which controllers it names and why it is untested on hardware.

Bernardo de Azevedo
Bernardo de Azevedo
· Updated
Felucca MPE: M-VAVE FM-1 pocket synthesizer beside the Yamaha CS-70M

Felucca MPE is a fork of the Felucca firmware for the M-VAVE FM-1, written by Stanley Gurvich of Emerald Moon Labs. It teaches the pocket synth to respond to MIDI Polyphonic Expression, so that each finger on a compatible controller can bend, press and slide its own note, and it adds two modulation sources that give every voice its own movement. The central question is practical: what does per-note expression change on a 27-key, roughly $70 instrument, and how far can a project that has so far run only in an emulator be trusted?

Table of Contents

What Felucca MPE adds to the FM-1

The M-VAVE FM-1 is a compact six-operator FM synth, and on a conventional MIDI setup a bend, a filter move or an LFO acts on every sounding note of a track at once. Felucca, the multi-engine firmware by Leo Kuroshita of Hügelton Instruments, already widened what the hardware can do. Felucca MPE keeps that base and changes how incoming MIDI is divided among voices.

According to the project’s README on GitHub, the work began on 13 October 2026 and sits on Felucca 1.6.1, with the same sounds, sequencer, effects and project files as the upstream release. The additions are a new MPE setting for MIDI IN, per-note pitch bend, two new modulation sources named PRES and TIMB for pressure and slide, and two sources that need no MPE controller at all, a voice LFO (VLFO) and DRIFT.

The framing matters for readers who already own a controller. MPE gives each note its own MIDI channel, which is what lets one finger glide while the rest of a chord holds still. A standard polyphonic keyboard cannot do that, and upstream Felucca offers no MPE option among its MIDI IN settings, which is the gap this fork fills.

Controllers and what the sources name

The README and the accompanying MPE.md name two controllers: a ROLI Seaboard RISE 49 and an Embodme Erae Touch (Mk 1). Both were played into the browser emulator, and the author’s verdict is that the result “feels like any other MPE synth.” That sentence describes the emulator, not the FM-1 itself.

MPE.md also gives setup notes for ROLI’s Seaboard and LUMI instruments in MPE mode through the ROLI Dashboard software, using the lower zone, 15 member channels and a bend range of 48. The LUMI is therefore mentioned as a setup target, but only the Seaboard RISE 49 and the Erae Touch are reported as actually played. No other controller is named, so claims about other MPE hardware, such as expressive pad or string controllers, would be guesswork until users report back.

Felucca MPE describes itself as not affiliated with M-VAVE, ROLI or Embodme.

How per-note expression maps to the sound

The pitch side is the most straightforward. Each member channel carries its own bend, which the firmware applies to one voice only. The default range is 48 semitones, matching the MPE convention, and the author states it can be raised to 96 through the RPN 0 message. Upstream Felucca limits bend to 24 semitones, so Felucca MPE also widens that ceiling for every MIDI IN setting. Two fingers on the same key become two separate voices, and a released note keeps its bend through its release tail.

Pressure and slide go through the existing modulation matrix on the MOD page. PRES reads the finger’s channel pressure, and TIMB reads the slide dimension, which MPE sends as CC74. Both are smoothed over roughly 3 milliseconds and belong to one voice each. They can be routed to the per-note destinations AMP, CUT, SHP and PITCH, with AMT setting the depth. On other destinations, such as pan or the effect sends, they take the value of the most recent note.

The documentation suggests two starting points. Routing PRES to AMP makes a note swell as the finger presses harder and stay silent at rest. Routing TIMB to CUT lets a slide along the key open the filter on that note alone. A small PRES-to-PITCH amount turns pressure into vibrato. Outside MPE mode, PRES falls back to the track’s channel aftertouch and TIMB reads zero.

For a pocket instrument this changes the playing feel considerably, since a finger can shape the filter or the tone-shaping control of one note while the rest of a chord stays put. The documented per-note destinations are limited to the four listed above, so the expression does not reach individual FM operators.

Voice LFO and DRIFT without a controller

The two modulation sources that do not depend on MPE are arguably the more broadly useful part, because they work on the FM-1’s own keys and on its sequencer under any MIDI IN setting.

VLFO takes the track’s LFO and reads it at a different point of its cycle for each of the eight voice slots, spaced 45 degrees apart. With an ordinary LFO, all notes of a chord move together. With VLFO routed to pitch at a low amount and slow rate, the notes drift against each other like an ensemble or a mild chorus. Routed to cutoff, it makes a chord shimmer note by note. VLFO is always bipolar, and with a sample-and-hold wave every voice receives the same value.

DRIFT is a slow random wander for each voice, independent of the LFO. According to the documentation, each voice picks a new random target about every three seconds and glides toward it at its own moment. On pitch, the README gives about 15 cents of deviation at AMT 1 and about 30 at AMT 2, which it likens to the drift of an old analog synth. Routed to cutoff, it makes the filter move slightly on every note.

The changelog also reports measurements from the author’s tests on a chord: with a pitch amount of 6, the plain LFO kept the notes within 6 cents of each other, while VLFO spread them by up to 158 cents. These figures come from the project’s own test runs and have not been independently reproduced.

Voice limits and what the FM-1 can sustain

The practical ceiling is the instrument’s polyphony. The FM-1 has eight voices shared across its four tracks, and MPE plays only the selected track, so that track gets all eight while the others fall silent. Each engine then has its own cap. According to the README, ANALOG, PHASE, LOFI, SAMPLE, SLICE, TRIO, WHEEL and NOISE allow eight fingers, FM6 allows six, VOICE four, and GRAIN and PHYS three. DRUM tracks play hits from notes and ignore finger bends.

The README explains the FM6 limit: each voice of that engine keeps a full six-operator state in the track’s engine memory, and six voices fill it. That is a notable constraint for readers who want the DX-style engine and an MPE controller together, since a seventh finger takes over an existing voice.

When voices run out, the firmware steals the oldest released note first, then the oldest held note, and never the lowest held note, so a bass line survives. MONO and LEGATO modes follow the latest finger, and UNISON stacks voices that bend together.

There is also a workaround for players without an MPE controller. Setting MIDI IN to CH1-4 and sending each note on its own channel from 1 to 4 lets each track bend independently, which gives four-finger behaviour when the same sound sits on all four tracks.

Who built it and where it sits among Felucca forks

The README credits Stanley Gurvich, Emerald Moon Labs, as the author and Leo Kuroshita of Hügelton Instruments as the author of Felucca. Gurvich’s personal site lists a handful of other tools, including a Relay FM-1 Editor, but that page does not describe Felucca MPE or its history, so little can be said about his background beyond the repository. Synthtopia credited a Gurvich video in an early October overview of FM-1 alternative firmware, in the section on upstream Felucca. That article does not mention MPE.

A second project covers similar ground from a different base. TamaMPE, published by the GitHub user tamabotmusic, is a fork of Melodee, another Felucca-derived firmware, and its About text calls it expressive MPE firmware for the FM-1. Its README, however, does not detail the MPE behaviour, and it does not say how it was tested beyond a build-machine test folder, so a feature comparison with Felucca MPE is not possible from the documents alone.

Upstream Felucca itself continues to move, although its direction has been unclear. The upstream README text that Felucca MPE reproduces carries a maintenance notice saying the maintainer is stepping back from active development, with occasional maintenance only. Upstream’s current README no longer shows that notice and describes the FM-1 firmware as being at version 1.7.0, a release that a release tracker also lists. Felucca MPE’s documentation states it is built on 1.6.1, so it lags that release at the time of writing, although its changelog shows one earlier merge of upstream changes (1.5.5 to 1.6.1).

Testing status, installation and limits

The documentation is direct about this. The README warns that everything so far ran in the browser emulator, which compiles the same C source to WebAssembly, and in Felucca’s host tests. It has not been flashed to an FM-1, and USB MIDI on the device and performance on the actual processor are untried. The 1.6.1 build has not been rechecked against the device’s size and audio-timing budgets.

A prebuilt package, felucca-mpe-1.6.1.fwsc, is included in the repository with a SHA-256 hash, and the README marks it as never installed on hardware. It uses Felucca’s package identity, so installers and editors will treat it as Felucca. No formal releases were listed on the repository page at the time of writing.

The author does provide size measurements, taken on an earlier 1.5.1-based build with the toolchain and not repeated for 1.6.1. They show about 480 bytes of extra flash and 528 bytes of RAM, with the audio interrupt cost rising from 38,871 to 41,218 against a stated budget of 37,502 and a 10 percent allowance. The headroom looks adequate on paper, but a figure from a different version is no substitute for a hardware run.

There are compatibility costs. MPE is stored in saved projects as MIDI IN value 6, which upstream Felucca does not recognise, so MIDI IN should be reset before moving a project to stock Felucca. Projects from the earlier 1.5.1-MPE builds stored the value differently and open as TRACK. Still on the author’s to-do list are the controller setting the zone size itself (RPN 6), the upper zone and release velocity.

Anyone tempted to flash it should first read the project’s warnings, keep M-VAVE’s updater or Felucca’s Return to official V15 option available, and follow the site’s safe install and recovery guide. Part of the reason such experiments are possible is described in Inside the M-VAVE FM-1.

Final perspective

Felucca MPE is a narrow, well-documented idea: give a very cheap synth the channel handling that expressive controllers assume, and add two cheap sources that make chords sound less rigid. The mapping is sensible, the documentation is candid, and the voice limits are stated plainly rather than hidden. For FM-1 owners who already have a Seaboard or an Erae Touch, it is a documented way to try that pairing, with TamaMPE the only comparable project named in this article.

The caveats are equally clear. No hardware run has been reported, no independent review had been published at the time of writing, and the controller list rests on one author’s emulator sessions. Playing is limited to one track at a time and to between three and eight simultaneous notes depending on the engine, and Felucca MPE is built on Felucca 1.6.1 while upstream has already moved to 1.7.0.

Until someone reports a flash and a playing session on a real FM-1, it is best treated as a promising prototype, with VLFO and DRIFT the parts most likely to be useful to players without an MPE controller.

Sources

  1. stanleygurvich/FeluccaMPE: README and repository page, GitHub
  2. Felucca MPE: MPE.md usage notes, GitHub
  3. hugelton/Felucca: upstream firmware repository, GitHub
  4. hugelton/Felucca v1.7.0 release notes, newreleases.io
  5. tamabotmusic/TamaMPE, GitHub
  6. Alt Firmware Options For The M-Wave FM-1 Pocket Synthesizer, Synthtopia
  7. Stanley Gurvich, author’s website
Bernardo de Azevedo
Written by
Bernardo de AzevedoFounder & Editor

His connection with music began at age 6, in the 1980s, when his father introduced him to Jean-Michel Jarre's Rendez-Vous on vinyl. He works professionally in the legal field, while synthesizers became his space for abstraction and creative exploration. Founder of The Synth Source.