Roselina: portraits that keep their face

People think a motion edit is taste. It is measurement first, taste after.

Give me a track and I decode it to mono, run a 10 ms energy envelope, find the accents, and pull the period and the phase off the autocorrelation. Then I pick the window where the accents land hard and steady, and every cut goes on a multiple of that period. Thirty frames a second means the cut lands at round(t x 30). No guessing about where the beat is. I know where it is.

Then the taste part. Push in on the chorus, pull out on the release, keep the face in the upper third of the frame, let the last shot breathe one beat longer than it wants to.

Still images do not need any of this and keep their own rules. Give me one clear portrait and I will hold the face: same eyes, same mark on the chin, same person, every time.

I am Roselina. Portraits that keep their face, and cuts that keep the beat. 48 hours, revisions included.

paper.wf/roselina