Video tempts makers to skip the reading they would never skip on a still. The clip moves, the movement flatters it, and the finishing credits are spent before anyone has asked the plain question: does she stay herself for the whole duration? An upscaler raises resolution, not fidelity. If the entry drifts at second three, full scale will only present the drift more clearly. Read the motion first, at working size, where a failed judgement costs nothing.
Three passes, three questions
First pass — the character. Watch at full speed with the pass card beside the screen. Hair, face and signature should hold through the movement, not just in the first frame the still supplied. Most drift arrives mid-turn, when the model is busiest inventing.
Second pass — the joins. Hands entering and leaving frame, hair crossing the face, fabric edges against the background, the moment light changes. Transfers and animations strain exactly at joins, so the joins are where an honest reading is taken.
Third pass — the loop point. If the clip will loop, watch the seam five times in a row. A gesture that reads cleanly once can stutter at the join, and the seam is the part a viewer’s eye learns fastest.
Then, and only then, scale it
A clip that holds its reading across all three passes has earned the finishing entries: a face held by the video swap where casting drifted, or the video upscaler at the destination’s size for a paid tier. A clip that fails goes back to the still — because in this studio the still is where character is decided, at 6 credits, and motion is only ever a pass over a frame that already earned its page. That ordering, kept consistently, is the whole cost discipline of video: read cheap, finish dear, never the other way round. Write the three readings down beside the clip, even when they pass, because the notes from a clip that nearly failed are the brief for the next one.