Short Frame Check

Can a moving caption cross the guide between two checked frames?

Yes, if its motion or changing shape extends beyond the checked positions. Our chosen overshooting path fits at both ends but crosses the bottom guide midway. A fixed-size caption translated along a straight path inside a fixed rectangular guide is a different case: fitting endpoints keep that geometric path inside.

Our chosen motion path

We built a synthetic 360 by 640 frame and a fixed caption rectangle 160 pixels wide and 40 pixels high. Its horizontal center stays at 180 pixels. We chose a vertical center function Y(t) equal to 360 plus 20 times t plus 240 times t times one minus t, where t runs from zero to one. This is an explicit experiment path, not an interpolation rule claimed for a video editor.

Use a chosen bottom guide at 416 pixels, matching the checker's default bottom percentage for this frame height. At t equals zero, Y is 360 pixels and the box bottom is 380. At t equals one, Y is 380 and the box bottom is 400. Both endpoints sit above the boundary, with clearances of 36 and 16 pixels respectively.

The middle frame gives a different result

At t equals one half, the linear part gives 370 pixels. The added excursion is 240 times 0.5 times 0.5 equals 60 pixels. The center is therefore 430 pixels and the bottom edge is 450. It crosses the chosen bottom guide by 450 minus 416 equals 34 pixels. Looking only at the endpoints misses that excursion.

Our script renders masks at the start, middle and end, then reads each occupied bounding rectangle back from the saved image geometry. Their vertical ranges are 340 through 380, 410 through 450 and 360 through 400 using half-open bounds. The horizontal range is 100 through 260 in all three. The measured boxes agree with the coordinate calculation, so the endpoints-fit and middle-fails result is reproducible with these inputs.

When endpoints really are enough geometrically

Remove the excursion term and keep only Y(t) equals 360 plus 20 times t. The center stays between 360 and 380, so the lower edge stays between 380 and 400. It cannot cross a boundary at 416. More generally, a fixed-size rectangle translated linearly between two positions inside a fixed rectangular guide remains inside: every edge coordinate is a weighted average of its two endpoint coordinates.

This argument requires the complete rectangle to fit at each endpoint. It also requires unchanged size and orientation, a fixed guide, and an actual straight interpolation path. A rotating, scaling or overshooting caption does not meet those assumptions. A software label such as ease is not enough to establish the path; inspect the resulting export.

What this experiment cannot certify

The checker previews the frame you select. This source trace and synthetic motion experiment do not establish automatic inspection of an entire clip, real subtitle detection or phone interface visibility. Review the movement and styling in the exported video, including intermediate moments where the caption moves farthest toward an excluded region. The displayed diagram samples our chosen path; it is not a universal worst-frame finder.

Start, middle and end of a deliberately chosen caption motion path
Synthetic fixed-box animation samples; the path is chosen for this experiment.

Questions and answers

Do two fitting screenshots approve every frame?

Not without knowing how the full caption geometry changes between them.

Does every animation overshoot?

No. Our experiment deliberately chooses a path with an excursion.

Does the tool scan a complete clip automatically?

No. Inspect different frames and review the exported animation.

Check an exported frame · Review different video moments

Visual walkthrough

Text-led walkthrough without audio.