cymfonic · humanizer
master track gives the chord. you decide how it plays — which chord tones, in what rhythm, at what velocity, in what voicing. capture a phrase once and it works in every key, against every chord, for the rest of the song.
vst3 · au · windows · mac · a harmonyengine plugin
phrase library
humanizer phrases are stored as degree patterns, not pitches. capture a four-bar arpeggio over Cmaj7 today. drop it into a song in F♯ minor tomorrow. it plays. it follows. it voices.
phrase library preview
humanisation
three sliders and a velocity range. that's the difference between programmed and performed.
questions
no — that's harmonyengine. you write roman numerals on a single master midi track, and harmonyengine broadcasts the chord state to every linked plugin in your daw. humanizer reads that state and turns it into voiced midi. think of harmonyengine as the harmonic score and humanizer as the part-writing.
yes. humanizer is a harmonyengine plugin — it has no chord data of its own. without a harmonyengine master track in the same daw session, humanizer passes midi through unchanged. the two are designed to be used together.
closed, open, drop 2, drop 3, with inversion. closed packs every chord tone into the tightest cluster. open spreads them an octave. drop 2 and drop 3 are jazz comping voicings that pull the second- or third-highest voice down an octave for a wider, more orchestral spread. inversion shifts which chord tone sits in the bass. you choose the voicing per region — different sections, different textures, all from the same piano roll input.
arm a region, play a pattern over the master's current chord, hit lock. the phrase is captured as a sequence of degrees — not absolute pitches — so it survives any chord change, key change, or modulation that comes after. tag it (arp, strum, bass, anything you like), save it as a .hephrase json file, recall it from any project on the same machine. the tag is the folder. the filesystem is the library.
no. humanizer is compositional scaffolding — it doesn't write the music, it makes the music you wrote playable. you decide which chord tones sound. you decide the rhythm. you decide the voicing. the plugin handles the mechanical translation of your harmonic intent into voiced midi, sample-accurate, in real time. we provide the bones. you bring the soul.
yes — it's a core use case. stack three humanizer instances on three tracks: one range-constrained to C2–C3 for bass, one to C3–C5 for comping, one to C5–C7 for pad. a single piano roll pattern voices itself across the full register. one harmonic source, full orchestral spread, no duplicated midi.
vst3 and au formats. windows and mac. tested in reaper and ableton live. logic pro support arrives with the au build. cubase, studio one, and bitwig are vst3 hosts and should work without ceremony — let us know if anything misbehaves.
four parameters, per region. timing jitter randomises note-on positions in a ±20ms range. length variation randomises note-off positions by 0–20% of the original length. velocity remap stretches incoming velocities into a per-region min / max window. fixed delay offsets every output event by a chosen amount for offsetting layered instances. all four are independent. all four save with the project.
updates
humanizer launches alongside harmonyengine v1.0 — vst3 and au, windows and mac, same release. sign up to hear when the closed beta opens, and to be first in when we ship.