Compare two videos
Put two clips next to each other, map where they differ, or measure the gap.
Your file
Drop a file here, or click to choose
It stays on this device. Nothing is sent anywhere.
The first file is the reference, this one is what it gets measured against.
Settings
Three ways to compare
- Side by side stacks the two clips in one frame. The honest way to judge a compression setting, because you see both at once rather than trusting your memory of the first.
- Difference subtracts one from the other, so anything identical goes black and anything changed lights up. This is how you find out exactly what an edit touched.
- Score computes a similarity figure across the whole clip, which is useful when the difference is too small to see but you still want a number.
Line them up first
Comparison only means anything if both clips start at the same moment. If one has a few extra frames at the head, trim it first, otherwise every frame will read as different.
This is the mistake that wastes the most time here. Video comparison is frame against frame, so a single frame of offset makes an identical pair of clips score as completely different, and the difference view lights up everywhere at once. If a result looks nonsensical, alignment is the first thing to check. Trim video in exact mode cuts on the frame you ask for, which is what alignment needs; fast mode snaps to a keyframe and will not do.
Reading the score
The metrics mode gives PSNR in decibels and SSIM from 0 to 1. Above 40 dB PSNR the clips are visually identical in any practical sense; 30 to 40 dB means a difference exists that most people will not notice without flipping between them; below 30 dB it is visible. SSIM tracks perception more closely, weighting structure over raw pixel error, and above 0.98 is generally indistinguishable while below 0.95 is not.
Both are averaged across the clip, which hides the thing you often care about. A video that scores well overall can still fall apart during its one fast-moving scene, because that is where the encoder ran out of bitrate. If a compression setting matters, compare the difficult section rather than the whole file.
What this is for
- Choosing a compression setting honestly. Export at quality 23 and quality 28, put them side by side, and decide whether the smaller file actually costs anything. Memory is unreliable; simultaneous viewing is not.
- Seeing what a platform did to your upload. Every service re-encodes what you give it. Compare your master against the downloaded version and the damage is visible rather than suspected.
- Confirming an operation was lossless. A stream copy should come back as an essentially perfect match. If it does not, something re-encoded when you believed it was copying.
- Finding what an edit touched, using the difference view, where everything unchanged goes black.
The clips are decoded and processed frame by frame, so this is slower than most tools here, and the cost scales with length and resolution. Compare a representative ten seconds rather than a full recording. For stills, compare images does the same job and includes an amplify control for making very small differences visible.
Common questions
The difference view is entirely black. Is it broken?
No, that means the two files are visually identical, which is the answer you wanted.
What score counts as good?
For a compression check, higher is closer to the original. Judge it relative to another setting on the same footage rather than against an absolute threshold.
Do the clips need the same resolution?
They are scaled to a common canvas first, so mismatched sizes still work, though a large mismatch makes the difference view less useful.