How to Compare a 4K Video to a 1080p Copy

You shot in 4K and delivered in 1080p, or a platform turned your 4K upload into a 1080p stream, and you want to know what changed. It is a reasonable question with two separate answers, and it helps to know which one a comparison gives you.

Two different questions

1. Is it still the same video? Same cut, same frames, same colour, nothing dropped or re-timed, no compression damage you would notice watching it.
2. How much detail did the lower resolution cost? The fine texture in hair, foliage and fabric that 4K has and 1080p cannot hold.

A side-by-side comparison is very good at the first and, by design, not the tool for the second.

How the comparison handles two sizes

A comparison has to measure both videos on the same grid, so the larger one is scaled to match. DiffALL’s video comparison goes further: it scores at a small, fixed size and samples moments across the clip, because that is what keeps a comparison of minutes of video fast and steady on a modest server.

Above 1080p there is one more step. Decoding two 4K videos at once needs more memory than the server has, so when either file is larger than 1080p, both are converted to 720p copies first, and the result says so:

Compared at 720p. These videos are 3840x2160, more than this server can hold in memory two at a time, so both were scaled down first.

We measured what that costs on a 4K pair, native against the 720p copies: a brightness change scored within 0.2 points either way, a drawn-in box within 0.1, and a slight blur read about three-quarters of a point closer at 720p. So visible changes score the same, and very fine detail can read a little closer than it would at full size.

4K is also slow to decode, so on a very large pair only the first few seconds may be compared; the note under the score says how much.

What that means for your 4K vs 1080p question

Because both sides are compared at a lower resolution than either file, the comparison will not show the detail that 1080p lost — that detail lives above the resolution it measures at. A high score here does not mean the 1080p copy is as sharp as the 4K original. It means it is the same video: nothing was cut, re-timed, re-coloured or visibly damaged.

That is still worth knowing. A downscale that went wrong usually goes wrong in ways the comparison does catch:

• Frames dropped or duplicated by a frame-rate conversion;
• A colour shift from a wrong colour-space conversion (HDR to SDR, or BT.2020 to BT.709);
• Blocking and smearing from too low a bitrate, especially in motion;
• A different cut — a trimmed start, a missing scene.

Checking the detail itself

For the second question, look at the 1080p file on its own with the video quality check. It reports sharpness, compression and bitrate for the file in front of it, judged for its own resolution. A 1080p file that reads Good on sharpness and bitrate has kept what 1080p can hold; one that reads Poor on bitrate was squeezed harder than the resolution alone explains. For what counts as enough, see is your video bitrate too low.

If you want the full clip compared

Export both versions at 1080p (the 4K master scaled down in your editor, against the delivered 1080p file) and compare those. They are within the size the server compares directly, so the whole clip is read at the scoring resolution with no extra conversion step.