A field is not half a frame. It is a different moment.

An interlaced frame is two pictures, taken one field period apart and combed together one line at a time. At 1080i25 that is 50 moments a second, 20 ms apart — and the file reports 25 frames, so half of what the camera captured is invisible to anything counting frames. The comb is not an artefact: it is the distance the subject travels between the two moments. Something crossing a 1920-pixel frame in two seconds moves 19.2 pixels, and those 19.2 pixels are the teeth.

The engine is free and needs no account. Which means the order of a processing chain is not a matter of taste. A step that resamples across lines — a scale, a thumbnail — reads the frame as one picture and averages both moments into every output line. The comb does not go away; it becomes a smear that is part of the picture, and no deinterlacer downstream can separate what has already been averaged. Scale then deinterlace and the footage is gone. Deinterlace then scale and it costs nothing. Same operations, same footage, different order. And it is invisible in every still frame, because interlacing halves the resolution only on the parts that move — 1080 lines on anything static, 540 on anything moving — which is why a chain tested on a locked-off shot has not been tested. Every lane ships a worked example that costs nothing. Writing the report needs credits.