How to Compare HDR and SDR Versions of a Video

The comparison everyone gets wrong first

You have an HDR master and an SDR version derived from it. You want to know whether the SDR conversion holds up.

The instinct is to compare them directly and read the similarity score. That number will be low, and it will not mean what you want it to mean. HDR and SDR use different transfer functions and different colour primaries — the same scene is encoded with genuinely different pixel values by design. A low score here measures the format difference, not the quality of your tone map.

The useful comparison is a different one: tone-map the HDR master to SDR yourself with a known-good setting, then compare that against the SDR version you are checking. Now both files are in the same colour space and any difference is attributable to the conversion, which is the thing you actually care about.

What goes wrong in tone mapping

Highlight clipping. HDR carries detail far above what SDR can represent. A conversion that maps too aggressively flattens bright areas into featureless white — skies, windows, specular highlights on skin and metal. The detail is not dimmed, it is gone.

Shadow crush. The opposite failure at the other end. Dark detail collapses to black, and anything shot low-key loses its shape.

Saturation shifts. Wide-gamut colour squeezed into Rec. 709 has to go somewhere. Done badly, saturated reds and deep blues either clip or drift in hue. Skin tones are the most noticeable and the least forgiving.

Inconsistency across a scene. Some tone mappers adapt per shot or per frame. When the adaptation moves during a cut or a camera move, the brightness pumps in a way that never appears in the HDR original.

That last one is the strongest argument for a per-frame comparison rather than spot-checking stills. A single frame looks fine; the problem only exists across time.

Running the comparison

Upload both SDR files — your reference conversion and the version under test — to DiffALL’s video comparison. You get an overall similarity score, a per-second chart, a difference heatmap, and the worst-matching moment identified for you.

The per-second chart is the important output. A flat line means your conversion behaved consistently. Dips mark the shots where it struggled, and those shots are where you should be looking.

The heatmap tells you where in the frame. Two patterns are worth recognising immediately:

• Bright regions in the sky, windows, or lamps — highlight handling. Your tone map is clipping where the HDR had detail.
• Bright regions in the darkest parts of the frame — shadow handling, crushing detail to black.

If the difference is spread evenly over the whole frame with no structure, that is usually a global brightness or gamma offset rather than a tone-mapping fault — easier to fix, and worth ruling out first.

Checking a specific scene type

Some content stresses a conversion far more than the rest. Test these deliberately:

• A bright sky with cloud detail — the classic highlight test.
• A night scene with practical lights — tests both ends at once.
• Faces under mixed lighting — where saturation errors are most visible and least acceptable.
• A hard cut between a dark and a bright shot — catches adaptive tone mappers pumping.

A twenty-second clip covering those cases tells you more than five minutes of average footage.

Comparing two tone-mapping settings

The same method works for choosing between conversions. Produce SDR versions with two different settings, compare them, and read the chart: it shows you exactly which shots the two treatments disagree about. Those shots are the ones to judge by eye. The rest are identical and do not need your attention.

This is much faster than watching both versions end to end and trying to remember what you saw.

A note on frame rates and lengths

Both files need to line up. Different frame rates, a different starting point, or a trimmed version against an untrimmed one will produce a comparison dominated by the misalignment rather than by the picture. Export both from the same in and out points at the same frame rate.

Try it

Compare two videos → — upload both versions and get a per-second quality chart, a difference heatmap, and the worst-matching moment. Compare SDR against SDR, not against the HDR master.