@jeffrey: "the pa in patiently should be pitched down i think · it feels too fast and abrupt, maybe it needs to be stretched to be a full bar · and can waiting start on bar 8:1 · and very needs to be two notes". The chart gains `shift`: N semitones on one unit, applied to her f0 AND to the chart note, so the pluck doubles the new pitch instead of the one she sang. "pa" goes down a scale step, F4 → D#4, so the phrase stops leaping up from very's A#3 before it descends — and it now holds bar 10 outright rather than two beats. "very" splits into ve · ry, and the split is not cosmetic: measured separately they are A#3 and G#3, two real notes that the single unit had been averaging into one. just and waiting take a bar each, so waiting starts on bar 8 beat 1. Also checked the sync he asked about, and the file is clean. The audio's first kick sits at 0.4910 s against a 0.490 s lead-in. For the video I measured the flashing kick cell sampled slightly off the beat, where it clears the playhead — mean +0.58 px, worst 1.2 px, i.e. +3.0 ms mean and 6.2 ms worst against a ±8.3 ms floor from 60 fps itself. Earlier attempts read −8 px only because the detector was finding the pink PLAYHEAD (4 px wide, and it crosses the kick lane) rather than the thing being measured. What likely reads as "a little early" is real but not a timing error: the roll scrolls through the pickup beat before the first kick sounds, so the eye sees motion before the ear gets anything.