Diagnose motion artifacts frame by frame

Use a triage row with file, scene, time, frame index, observed defect, expected appearance, likely checks, change made and retest evidence. An illustrative entry is “Heading accent disappears for one frame near its entrance.” That describes a symptom. “Font bug” is a hypothesis and should not replace the observation.

Save the frame before, the affected frame and the frame after. If the issue only appears in playback, also preserve a short segment around it. Confirm whether it exists in the encoded file or only in one player before changing the source. A player-specific observation may need a different investigation.

For clipping, inspect text bounds, masks, overflow regions and maximum motion excursion. For jagged edges or shimmering thin lines, inspect scale changes, raster assets, subpixel placement and the final delivery size. For a broken mask, inspect its coordinate system and the frame when it becomes active.

For a one-frame jump, compare scene start/end conditions, rounded frame calculations and asset readiness. These are investigation paths, not guaranteed causes. Change one plausible cause at a time so that a successful retest can be connected to an actual correction rather than a collection of unrelated edits.

Suppose a capital with an accent is fully visible at rest but clips while sliding through a narrow text mask. The proposed correction expands the mask to include the glyph’s full bounds throughout the motion. Preserve the original reference frame and compare the revised frames at the same point in the sequence.

Do not call the example fixed merely because the source mask is larger. Render the affected interval, inspect adjacent frames and play it at ordinary speed. If timing changed, compare the equivalent visual state as well as the old timestamp. Record both observations so the evidence remains understandable.

After correcting a shared text mask, inspect other scenes using that mask and the longest relevant labels. After changing a global scale or renderer setting, broader output review may be justified. A local copy correction does not automatically require unrelated production experiments, but shared dependencies do require attention.

Record the new encoded file identifier, inspection frames and remaining concerns. A clean still frame does not prove smooth movement, and a smooth preview does not prove the final encode. Keep both real-time playback and frame-level evidence in the review reel. Leave unresolved hypotheses visible so the next producer does not mistake them for established causes.