How to use Hum · The manual

At a glance MT-03

Hum sits on your mic track and turns a hummed, sung or whistled line into MIDI, snapped to your key. The audio passes through untouched, so you can monitor your voice while it writes notes.

The editor teaches instead of hiding: your raw pitch is a thin line, the note it emits is a block on the scale, and a meter shows exactly why the gate did or did not fire.

The pipeline Signal, not screenshot

voice in ─→ 70 Hz high-pass ─→ YIN pitch ─→ note gate ─→ scale snap ─→ MIDI out
            (analysis copy)     (f0 + conf)   (adaptive    (KEY + SCALE)
                                             floor)
voice in ────────────────────────────────────→ audio out  (dry)

Set up From insert to first sound

01 Insert Hum on your microphone track. The dry audio passes straight through; the MIDI comes out the side.
02 Set KEY and SCALE to your song so the line snaps in tune. Leave SCALE on Chromatic to keep every semitone.
03 Hum one line, not a chord. Watch the state chip: it reads LISTENING, TOO QUIET or UNSURE, then names the note when one commits.
04 If nothing triggers, look at the gate meter: if your level bar does not reach the clay tick, raise SENSITIVITY or get closer to the mic.

The controls Every parameter

Snap

KEY C … B: The root of the scale the line is snapped to.
SCALE Chromatic · Major · Minor · Dorian · Mixolydian: The snap scale, Major by default. Chromatic keeps every semitone; the rest pull each note to the nearest scale tone, preferring the note below.
OCT -2 … +2 oct: Transposes the emitted note in whole octaves after snapping.

The gate

SENSITIVITY 0 … 1: The gate’s eagerness, not a level, shipping at the 0.5 midpoint. Low is deliberate: a note needs about 24 dB over the room’s noise floor and strong pitch confidence. High is eager: around 10 dB of margin and a looser gate. Because it works against an adaptive floor, whether a note fires means the same thing on every microphone.
TIGHTNESS 0 … 1: Pulls note onsets forward onto the host’s next 1/16 while the transport runs, always by waiting for the grid, never by jumping back. 0 leaves your timing free; 1 lands notes on the line. It is inert when the host is stopped.

Commit

LATENCY MODE Live · Accurate: Live, the default, answers fastest and reports zero latency, taking a little onset jitter. Accurate waits for a settled pitch before committing a note and reports its analysis window to the host so the MIDI lands in the right place. One honest trade in Accurate: the dry audio stays zero-latency, so a compensating host slides your monitored voice slightly early against the other tracks.

The display

Read-only, and the point of the plugin. It shows every decision as it happens.

PITCH LANE readout: The last few seconds, scrolling. The paper rungs are the in-scale semitones (the root drawn stronger), the thin line is your raw voice, and the clay block is the note Hum emits. The gap between line and block is the snap.
GATE METER readout: Your input level as the bar, the adaptive noise floor as the hatched region, and the trigger as the clay tick. If the bar does not reach the tick, that is why nothing fired.
STATE CHIP readout: The gate’s verdict in words: LISTENING near silence, TOO QUIET under the trigger, UNSURE when it is loud but not pitched, a note name while an onset builds, and the heard note snapping to the emitted note once it is out.

First five minutes A short walkthrough

01 Insert Hum on the mic track and set KEY and SCALE to the song.
02 Hum a steady note and watch the block appear on the scale rung.
03 Hum a short line, leaning into each note a little: an onset needs a few agreeing pitch readings before it commits.
04 If it misses quiet notes, raise SENSITIVITY and check the gate meter; if it fires on room noise, lower it.
05 Turn TIGHTNESS up with the transport running to pull your onsets onto the grid, then record the MIDI.

Recipes Named workflows

Catch a topline

01 Set SCALE to the song’s key and leave LATENCY MODE on Live.
02 Hum the melody in one pass, one note at a time.
03 Nudge SENSITIVITY until the quiet ends of phrases still register, then record.

A tight, in-time part

01 Switch LATENCY MODE to Accurate for steadier pitch, and start the transport.
02 Raise TIGHTNESS so onsets pull to the 1/16 grid.
03 Sing the part; the MIDI lands close to the beat and needs less editing.

A bassline by hum

01 Keep SCALE on the song key and set OCT down if you hummed above the target register.
02 Hum the line low and even, holding each note through its length.
03 Let a real slide retrigger the note where you want a new one; steady vibrato stays a single note.

Good to know The fine print

Hum is monophonic by design, from 60 to 1200 Hz. Hum one line at a time; it will not detect chords, and it does not promise polyphony.
An onset needs a few agreeing pitch readings, and a note sounds for at least 60 ms before it can end, so very fast runs may drop or merge notes. This is the gate refusing to guess rather than a fault.
Vibrato holds as one note; a deliberate slide of more than about four fifths of a semitone retriggers. That is how a scoop stays one note and a real step becomes two.
Detection is after the fact, so a note lands roughly 35 to 60 ms behind your voice. Live takes the jitter for speed; Accurate reports its window so the recorded MIDI sits in time.
Notes carry velocity from how loudly you sang, breath to shout. Unlike the gate, velocity reads absolute level, so it does follow your interface gain.
Incoming MIDI passes through byte for byte with Hum’s notes added on channel 1, and a transport jump or reset never leaves a note stuck.
Every plugin in the suite maps the same way: click the MIDI chip in the header (it reads MIDI, MIDI · N, or MIDI · OFF) and the editor dims into learn mode with a balloon on every mappable control, plus ENABLE MIDI per instance, ALL PARAMETERS..., CLEAR ALL and DONE (Esc works too). One CC drives one parameter, on any channel, saved the moment you map it; switches and short lists step relative to the knob. The window also resizes from its corner.