Innertia Audio
03 — Virtual analogue synthesizer

Jove

Polysynth with a patchbay

Most synths give you a fixed voice and a modulation matrix to wiggle it with. Jove lets you change what the voice is — patch an oscillator into another one’s frequency, run the filters in series, or close a feedback loop that will not blow up.

The Jove voice page: five oscillators with their waveform displays, sub and noise, the mixer, and the filter and voicing sections
Type
Virtual analogue synthesizer
Formats
VST3 · AU · Standalone
Systems
macOS · Windows · Linux
Version
v1.1.1 · 2026-07-19
Licence
GPL-3.0

Measured figures

11 × 11
Patch points, bipolar, rewiring the audio path rather than offering a routing menu.
1 sample
Of delay in the patchbay feedback path, which is why a patched loop cannot blow up.
32 slots
Of modulation, drawn as arcs and moving dots on the knobs they are pulling.
±1 dB
How closely the 140 factory patches are loudness-matched across the whole bank.
Oversampling of the entire voice path on the highest quality setting.

The patchbay is the point

The old EMS synths had a pin matrix instead of a signal path. You decided what connected to what, and the instrument was whatever those pins said it was. Jove has eleven source nodes crossing eleven destination buses, with a gain at every crossing that can go negative.

This is not a routing menu with a nicer skin. It rewires the actual audio path. Frequency-modulate any oscillator from any node in the instrument. Chain the two filters in series when the patch wants twenty-four more decibels of slope, or run them in parallel when it wants two characters at once. Patch the output of something back into its own input — which is safe, because the loop is delayed by exactly one sample, so it can howl but it cannot explode.

Half the patches I keep came from putting a pin somewhere I had not intended and leaving it there.

The patchbay: eleven sources crossing eleven destinations, with bipolar gains
The patchbayThe patchbay: eleven sources crossing eleven destinations, with bipolar gains

Five oscillators and seven filters

Each oscillator morphs continuously from sine through triangle and saw to pulse with variable width, band-limited so the top octave stays clean — or switches to wavetable mode and scans a bank of sixty-four tables. Each has its own footage, detune, bit-crush and sample-rate reduction, so you can dirty one voice inside an otherwise clean patch.

Underneath there is a sub oscillator an octave or two down, noise, ring modulation, oscillator-to-oscillator FM, and hard and soft sync. Unison stacks up to seven voices per note with detune and stereo spread, on top of sixteen-voice polyphony.

The filters are a Moog-style twenty-four decibel ladder, a state-variable in low-pass, high-pass, band-pass and notch, a low-pass gate for plucked and struck behaviour, and a Steiner-Parker. They are cutoff-matched to each other, which sounds like a small thing and is not: it means switching model while auditioning changes the character without changing the brightness, so you are comparing filters rather than comparing levels.

Modulation you can see moving

Three envelopes, three LFOs with six shapes each — tempo-syncable, with fade and delay, per voice or global — and four sixteen-step sequencers that run in melodic or curve mode with swing and direction control. All of them are sources in a thirty-two slot matrix, edited as a grid you can read at a glance.

The part I use most is that the matrix draws itself onto the instrument. Any knob that is being modulated shows the range as an arc around it and a dot moving live inside that arc. You can see the filter envelope pulling on the cutoff while you play, which turns editing from an act of imagination into an act of looking.

The modulation matrix, lit while it moves
The matrixThe modulation matrix, lit while it moves

Getting from nothing to a sound

There are a hundred and forty factory patches in twelve categories, and they are loudness-matched across the entire bank to within a decibel — measured, not eyeballed. Browsing the bank does not mean riding the output fader, and a patch that seems better is better rather than just louder. Each one still has a trim if you disagree.

Audition mode plays a demo phrase appropriate to the category at the host tempo, so you can go through a bank with your hands off the keyboard. And when nothing in it is right, DICE builds a new patch from a musical archetype rather than randomising every parameter into noise, while VARY perturbs what you already have. A/B compare, copy between slots, revert to the stored preset, and an indicator when there are unsaved edits.

There is an arpeggiator with twelve modes, ratcheting, swing, latch and host-transport lock. And three quality settings — Eco, HQ and Ultra — that oversample the whole voice path at one, two or four times, so you can write with a light CPU load and render with a heavy one.

Everything it does

A patchbay, not a routing menu
Eleven sources across eleven destinations with bipolar gains, rewiring the actual audio path. FM any oscillator from any node, chain the filters, or close a feedback loop that is delayed one sample so it cannot blow up.
Five oscillators that can each be different
Continuous morph from sine to pulse, or a 64-table wavetable scan, with per-oscillator footage, detune, bit-crush and sample-rate reduction — so one voice can be filthy inside a clean patch.
Seven filters, level-matched
A ladder, a state-variable in four modes, a low-pass gate and a Steiner-Parker, matched at the cutoff so switching model changes the character and not the loudness.
Modulation drawn on the controls
A 32-slot matrix edited as a grid, with every modulated knob showing its range as an arc and a live dot moving inside it. You watch the envelope pull the cutoff instead of imagining it.
A bank you can browse by ear
140 patches loudness-matched across the whole set to within a decibel, with an audition mode that plays a category-appropriate phrase at your host tempo.
Somewhere to start when you are stuck
DICE builds a new patch from a musical archetype rather than randomising into noise; VARY perturbs the one you have. A/B compare and revert-to-preset throughout.
Written light, rendered heavy
Eco, HQ and Ultra oversample the entire voice path at 1×, 2× or 4×, so arrangement and bounce can have different CPU budgets.

Specifications

TypeVirtual analogue synthesizer
Oscillators5 morphing or wavetable, plus sub, noise and ring modulation
Wavetables64-table bank, scannable
Filters2 filters, 7 models, single, serial or parallel, cutoff-matched
Modulation3 envelopes, 3 LFOs, 4 step sequencers into a 32-slot matrix
Patchbay11 sources × 11 destinations, bipolar, one-sample-delayed feedback
PolyphonyUp to 16 voices; unison to 7 per note with detune and spread
ExpressionFull MPE — per-note bend, pressure and timbre
Arpeggiator12 modes with ratcheting, swing, latch and transport lock
EffectsDrive (5 models), multiband saturation, chorus, phaser, delay, reverb
QualityEco, HQ and Ultra — 1×, 2× or 4× oversampling of the voice path
Presets140 in 12 categories, loudness-matched to ±1 dB
FormatsVST3 · AU · Standalone
SystemsmacOS · Windows · Linux
Current releasev1.1.1, 2026-07-19

Screens

Captured from the running plugin.

Four sixteen-step sequencers in melodic and curve mode
Sequencers
Drive, multiband saturation, chorus, phaser and delay
Effects

Download

Current release v1.1.1, 2026-07-19. macOS builds are not notarised yet — right-click the plugin and choose Open the first time.