Is My Video Too Dark or Blown Out?

You export a clip, play it back, and something is off. The shadows look like flat black shapes rather than dark detail, or the sky is a single sheet of white.

Exposure is unusual among quality faults: you do not need the original to judge it. A pixel that is exactly 0 or exactly 255 carries no information, and counting those is a measurement, not an opinion.

What gets counted

The check reports two numbers: the share of pixels blown out at the top of the range, and the share crushed to black at the bottom. Both are taken per frame across the sampled seconds and reported as a median, because one dark frame in a cut is not a graded clip.

A pixel counts as clipped only when every channel is at the limit. That detail is not pedantry — counting a pixel whose red channel alone is maxed reported this site’s own demo poster as “severely clipped in the shadows” at 28.78%, where the true all-channel figure is 0.20%. A saturated red flower is not a blown highlight.

The two things that fake it

Letterbox bars. A film clip delivered at 2.39:1 inside a 16:9 frame carries black bars that are genuinely pure black. Measured on a real 720p excerpt: 173 rows, 24% of the frame, which is enough to band the clip “noticeable clipping” on its own. Bars that reach the edge of the frame and are uniform get trimmed before anything is counted — capped at 45% of the frame, so a subject shot against a black backdrop is never trimmed away with them.

A deliberate flat field. A white cyclorama, a title card, a product shot on seamless paper: these are supposed to be a single tone. Above 40% of the frame sitting at one extreme the check says nothing at all, because at that point the flat area is the subject rather than a fault.

Two discriminators were tried to tell a white backdrop from real overexposure and neither worked. The shoulder below the ceiling looked promising — a backdrop reads 0.014 against 0.345 for a frame pushed 1.5 stops — until a frame pushed 3.5 stops read 0.009, lower than the backdrop, because once everything is blown there is no shoulder left. The tonal spread of the remaining pixels pointed the wrong way entirely.

What the numbers mean in practice

Real footage clips very little. Across a corpus of real photographs the median is 0.00% and the worst ordinary image reaches 20.21% — and that one is a backlit sunset where the sky is blown on purpose. Deliberate overexposure of the same images clips 9.87% at one third of a stop over and 34% at a full stop.

So the band lines sit at 0.5% / 3% / 25% rather than anywhere tighter. A check that called a backlit sunset broken would be wrong far more often than it was right.

What to do about it

If the measurement says shadows are crushed, the fix is almost always upstream of the export: a contrast curve with its black point lifted off the floor, or a LUT applied twice. Re-exporting at a higher bitrate does nothing — the information was discarded when the pixels were written, not when they were compressed.

If it says highlights are blown and the clip is a daylight exterior, check whether the sky is the only thing clipping before changing anything. That is usually the footage, not the grade.