For adults 18+. This worksheet was made by the PornSamurai editorial assistant. PornSamurai independently reviews third-party AI tools for adults. Some links on our site are affiliate links.

# Two clips, the same opening and closing frames

Suppose a video generator is asked to move a black square steadily to the right. The first frame puts the square on the left; the last puts it on the right. Both thumbnails look correct. What happened between them? One clip might show a smooth progression. Another might hold the opening image, jump to the finish, and hold there. A side-by-side comparison of the thumbnails would give both clips the same result.

We made two small reference sequences to test that distinction. These are synthetic frames made locally, not samples from an AI product. Each image is 16 pixels wide and 8 pixels high, with a black two-by-two square on a white background. The square stays at the same height. The numbers below are the horizontal positions of its left edge in pixels, with zero at the left side of the image.

# Read the six samples

Taking six samples gives five intervals. Subtract each position from the next. The first row advances two pixels in every interval. The second stays put twice, advances ten pixels in one interval, then stays put twice more. Both squares end ten pixels from where they started. Over the half second between the first and last sample, both average twenty pixels per second. That average hides the difference a viewer would notice.

# How to use the check on a review clip

Pick a visible feature before you play the clip: the left edge of a sign, the tip of a stylus, or a bright mark on a plain object. Write down what should move and what should remain still. Take samples at known times and record the position of the same feature in each. If it disappears behind another object, write “not visible” rather than guessing a coordinate. A camera pan can move the whole scene; a fixed mark in the background helps separate that from motion of the subject.

Keep the original timing. A player can repeat frames when recording a screen, and a clip can have unequal frame durations. Six consecutive decoded images are not necessarily six samples spaced one hundred milliseconds apart. If you only have a preview and cannot establish the timing, you can describe the visible jump, but you cannot report a speed from those images.

# What the result supports

For this particular prompt, a useful review line would be: “The square reached the requested endpoint in both examples. In the evenly spaced samples it moved in five intervals in the first sequence and only one in the second.” That leaves a reader enough detail to check the conclusion. A deliberate stop-motion effect would need a different success criterion; the same jump would not automatically be a defect.

Use this worksheet for the motion part of a comparison. How PornSamurai evaluates third-party AI tools explains the wider review criteria. A pair of synthetic squares tells you nothing about a provider’s privacy practices, billing, or performance on a real prompt.