Generative sequencer modules usually ship one idea. The Turing Machine is a shift register. Euclidean modules do one rhythmic algorithm. Marbles is a random generator with a deterministic streak. The focus is a virtue — you learn the one thing deeply, and the module has a personality.
Kaona’s B.A.C.H. goes the other way: twelve compositional systems in one panel.
What it is
A four-voice real-time composition generator, with 12 different compositional systems including:
- Fugue and canon
- Markov chains
- Cellular automata
- Fractals
- Generative rhythmic structures
It drives Eurorack and MIDI synthesizers from Gate, Pitch and Mod outputs, with:
- 8 CV inputs
- 128 presets on an SD card
- Pitch outputs across eight octaves (0–8V)
€594, with a 15% early-bird discount, first delivery mid-October 2026.
Why “fugue” in a module is a stronger claim than “Markov chain”
The twelve systems are not equivalent, and the difference tells you what the module is actually attempting.
Markov chains, cellular automata and fractals are generators. They produce sequences from a rule and a state. They have no notion of harmony, of voice, or of what the other three voices are doing. They are abundant in modular — a Markov chain is a weighted state machine, and several modules do it.
Fugue and canon are contrapuntal forms, and they are relational. A canon means voice two plays voice one’s melody, displaced in time and possibly in pitch, in inversion, in augmentation. A fugue means a subject is stated, answered in another voice at the fifth, developed through episodes, and subjected to stretto and inversion — under voice-leading constraints about what intervals may sound together and how they may move.
Implementing those requires the module to know about the other voices. A four-voice fugue generator is not four independent sequencers; it is one system producing four parts that are about each other. That is a genuinely harder thing to build in a Eurorack panel than another probability engine, and it is what the name is claiming.
Whether it delivers a convincing fugue or a plausible gesture toward one is the whole question, and a €594 pre-order is not an answer. But the attempt is in a direction almost nobody in modular goes, because modular’s culture is overwhelmingly about texture and process rather than counterpoint.
The eight-octave pitch output is the spec that tells you it is serious
0–8V across eight octaves means it is using the 1V/octave standard across its full useful range. That sounds trivial and it is the thing that separates a module that can play music from one that can play a riff.
Four-voice counterpoint needs range — a bass voice and a soprano voice several octaves apart, moving independently. A module with a two- or three-octave output can produce a sequence; it cannot produce four-part writing, because the voices will be crowded into the same tessitura and the counterpoint will be unintelligible.
Eight CV inputs is the other number that matters: it means the twelve systems’ parameters are modulatable from the rest of the case, which is the whole reason to do this in Eurorack rather than in software. A fugue whose subject length is being modulated by an LFO is not something a DAW plugin is set up to do.
128 presets on SD is the unglamorous one and it is what makes it performable. A twelve-system generative module has a large state space; without recall you would spend the set finding the patch rather than playing it. (Compare the Adamsynths Warthog we covered last week, which solves the same recall problem for an entire analogue system.)
The lineage
Algorithmic composition in hardware has a long and distinguished history that modular mostly ignores. Iannis Xenakis built stochastic composition from Poisson distributions and Markov chains in the 1950s and then into UPIC. Gottfried Michael Koenig’s Project 1 and Project 2 at Utrecht were serial and aleatoric composition programs. Laurie Spiegel’s Music Mouse was an intelligent instrument that enforced harmonic rules while you played. David Cope’s EMI analysed and generated in composers’ styles.
A module that offers fugue generation alongside cellular automata is standing in that tradition rather than the Buchla/Serge one, and it is a tradition with much more to say about pitch than modular synthesis usually does.
What to be careful about
Twelve systems in one panel risks being shallow in twelve ways. The focused modules earn their reputations because a single well-implemented algorithm with good controls is deep. A menu of twelve could be a Swiss Army knife — and a module whose systems live behind a small screen and a preset system is a different instrument from one where each control does one thing.
Pre-order, with first delivery mid-October. That is close, which is a good sign, and it is still a pre-order.
And “composition generator” is a claim with a long history of disappointment. The interesting question is not whether it produces a correct fugue — software has done that since the 1980s — but whether what comes out is worth listening to and worth playing with. Watch for demos from people who are not the manufacturer.