Double turns one recording into a vocal that sounds like several sung takes. Instead of a copy with chorus you get four virtual layers (L1, R1, L2, R2) — the full stack is on by default — each with its own timing, pitch, pan, level and formants. The outer pair L1/R1 plays louder and wider (±90), while the inner pair L2/R2 has a slightly shifted formant (±0.3 semitone) and a different tuning, so the four voices don't melt into one. Humanize listens to the syllables and gives every one of them a fresh, small offset in timing, tuning and level — like a second singer who comes in a touch early one time and a touch late the next. The change happens right before the syllable, so consonants stay sharp and the double never flanges or wobbles like a cheap chorus. The original stays untouched in the centre with no delay — the plug-in adds zero latency. Auto Gain turns the output down by exactly the power the doubles add, so the double doesn't "win" just by being louder — you compare width and thickness, not decibels (switch it off on an AUX send). Clarity tames the “s”, “t” and “k” in the doubles so the lyrics stay clear, while Tone, Saturation and Low Cut give the doubles their own character. Mono Safe and Width decide how much of the double survives in mono. Intervals up to ±12 semitones with preserved formants let you add an octave-down layer under a hook or fifths in the backing vocals. The display shows the stage: layers as points in pan and time (drag them with the mouse), the live position of every layer, the humanize range and the comb-filter zone. Next to it: a goniometer, correlation and mono loss — you can see right away whether the double will survive on a phone and in a club. Summed to mono, the full stack loses about 1 LU and doesn't turn into a comb filter. Built for rap, trap, R&B and pop: from a subtle double to wide hooks, ad-libs and backing vocals.