Keeping thirteen shots in one grade, and spotting the one that drifted

Grading shot by shot, each in isolation, produces thirteen images that are each defensible. Played in sequence or laid out in a grid they read as inconsistent, because your reference drifted between them — you were matching each shot to your idea of correct, and that idea moved.

The audience does not see thirteen images. They see the differences between them, and a cut where the skin warms slightly is more noticeable than either shot being marginally off in the same direction.

Matching is therefore a separate operation from grading, and it needs its own pass.

Choose a reference shot — typical rather than best, in the dominant lighting — and grade it fully. This is now the standard.

Grab it as a still. For every other shot, display the still alongside and adjust until they sit together. You are not asking whether the shot looks good; you are asking whether it belongs with the reference.

The distinction matters. 'Does this look right' produces drift because the answer depends on what you looked at last. 'Does this match that' has a stable answer.

Displaying two images side by side or as a wipe defeats adaptation, because both are present at once. Toggling between them does not — by the time the second appears your eye has adjusted.

Back it with scopes. Compare the reference and the shot on a parade and the differences are numerical rather than perceptual. Skin in particular is easier to match on a vectorscope than by eye, because the target position is visible.

Match in a fixed order: black point, white point, mid-tone brightness, then colour. Doing colour first means redoing it after every luminance change.

Lay every shot out as thumbnails in one view. Differences that were invisible at full size become obvious — one frame slightly warmer, one slightly brighter, one with more contrast.

This works because at small size you stop reading content and start reading tone, which is exactly what you are trying to assess.

Do it repeatedly during the matching pass, not only at the end. Catching a drifting shot at number six is cheaper than discovering at thirteen that shots six through nine went one way together.

Sometimes a shot was captured under conditions too different to match fully — a different time of day, a different light source, a different camera. Pushing it all the way produces an image that is visibly corrected: strange skin, blocked shadows, a cast in the neutrals.

In that case, move it as far as it goes cleanly and stop. A shot that is slightly different but clean reads as a shot; one that is matched and damaged reads as a mistake.

Where the difference is still too large, the honest answers are to reframe its role in the sequence or to replace it.

When every shot is matched, play the whole thing without stopping. Note where your eye catches. Do not fix anything during the pass — stopping to adjust reintroduces the isolation problem you just solved.

Then do a second pass after a break. Fresh eyes catch different things, and the second list is usually shorter and more accurate than the first.

Fix only what appeared on both lists. Items that showed up once were probably you looking for problems rather than seeing them.