Why a fade between two stills never looks like the scene actually moved

I keep running into the same cut: two photographs that belong together, but nothing in between.

A before-and-after of a room. A wide shot and a close-up of the same person. A daylight frame and a night frame of the same street. In a regular timeline this is usually a cross-dissolve. It is fast. It is also obviously a still sitting on top of another still. The camera did not move. The light did not change. You just watched the opacity slider.

For a while I treated that as a skill issue. Longer dissolve. A zoom. A warp in After Effects. Those tricks help when you already have video. They do not invent motion that was never filmed. If all you have is two JPEGs, the editor is guessing with geometry, not with the scene.

What I wanted was narrower than “make a video.” I wanted the in-between: a few seconds that start on frame A and arrive on frame B without looking like a PowerPoint. That is a different job from color, captions, or a full edit.

The workflow that finally stuck is small. Export two stills. Treat one as the first frame and one as the last frame. Let a model fill the motion, then drop the clip on the timeline like any other B-roll. Duration stays short — a few seconds is enough. If the pair is a bad match (different lens, different subject, heavy text in the frame), the result still looks fake. The model does not save a bad pair. It only saves the hours of keyframing when the pair is honest.

I also stopped opening a desktop app for the test pass. Guest upload in the browser is enough to see whether the motion is usable before I commit to an export. Signing in is only for the actual generate-and-download step.

If you want to try that first-and-last-frame pass on a pair you already have, you can generate a transition between two images in the browser and judge the motion yourself.

That is the whole loop I use now: pick two stills that already share a scene, generate a short clip, keep it or throw it away, then edit as usual. I would rather fail in twenty seconds than spend an afternoon on a dissolve I already know I will mute in the final cut.

Curious what pairs other people would try first — same location at two times of day, or two angles of one object?