An ear test for a voice. Sharper like this, or like this? Your preference designs it — nothing to configure.
Two candidates per round. Play both, keep the one you prefer. Nothing to configure — your ear steers the search.
Register × warmth. The glow is where your voice most likely sits — it tightens and brightens with every choice you make.
Hear it against the untouched source. If a value drifted from zero, you shaped it by ear alone.
How this works. It works like an eye test at the optician — “sharper like this, or like this?” — only for a voice. One recorded line is shaped in real time by four audio controls — detune (register), a low/high shelf balance (warmth), a presence peak near 2.6 kHz, and a short reverb (air). Each round draws two candidates from a search window and asks only which you prefer. Your choice pulls the window toward the sample you kept and shrinks it, so the two candidates grow closer round by round until they converge on one voice. Every value shown is the actual signal setting applied to the sample. The audio never leaves this page.