How to Tell if Two Videos Are the Same (or What Changed)

Why file size and checksums can’t answer it

You have two video files — maybe one came back from a client, one was re-uploaded to a shared drive, or you found near-duplicates in a media library — and you need to know: are these the same video?

The obvious checks mislead:

File size changes with every re-encode, even when the footage is untouched. The same clip exported at a different bitrate can be half the size.
Checksums (MD5/SHA) only prove byte identity. Re-encoding, trimming a black frame, remuxing MP4→MKV, or even re-saving with different metadata changes every byte while the picture stays the same. A checksum mismatch tells you nothing about the content.
Watching both works for gross differences, but not for a re-encode that softened detail, a colour shift, or three seconds trimmed from the middle.

The reliable way is to compare the content: decode both videos and measure how similar the frames actually are.

The three possible answers

Comparing two videos properly gives you one of three verdicts:

Byte-identical — same file, just renamed or copied. (A checksum settles this case in seconds; if the checksums match you’re done.)
Same footage, different encode — one is a compressed, resized, or converted copy of the other. Frame content matches almost everywhere; similarity scores stay high with small uniform loss from compression.
Different content — edits, cuts, different takes, added overlays, or entirely different footage. Similarity collapses at specific moments — and those moments are exactly what you want to see.

Check it in one upload

DiffALL’s video comparison tool decodes both files and compares them second by second:

Upload both videos — formats don’t need to match; MP4 vs MKV vs WebM is fine, and different resolutions are scaled to a common size before comparing.
It aligns the timelines automatically. If one file starts up to a couple of seconds late — a trimmed intro, an added logo card — the offset is detected and corrected before scoring, so a shifted copy isn’t misreported as “different”.
Read the verdict. An overall similarity score, a per-second SSIM chart, and the single worst-matching moment rendered with a difference heatmap.

Reading the result

What you see	What it means
Similarity 97–100%, flat chart	Same footage — at most a lossless or near-lossless re-encode.
Similarity 90–97%, uniformly lower chart	Same footage, visibly re-compressed or resized. The loss is everywhere, not in one place.
High similarity with a few sharp dips	Same video with localised changes — an edited scene, an inserted frame, a watermark or overlay appearing. Check the flagged seconds.
Score collapses from a specific second onward	One video is trimmed or has extra content — the timelines drift apart at that point.
Low similarity throughout	Different footage.

Common situations this settles

Duplicate hunting: a media folder with final.mp4, final_v2.mp4, and final_export_new.mp4 — find out which are actually the same and which copy kept the most quality.
Re-upload verification: confirm the video a platform or vendor returned to you is your footage, not a re-encoded or altered version.
Delivery check: the file your editor sent back — is it the approved cut, or did something change? The per-second chart shows every edited moment, in the spirit of a frame-by-frame comparison.
Archive integrity: verify a converted archival copy is faithful to the master without watching hours of footage.

The bottom line

“Are these two videos the same?” has a measurable answer. Checksums only cover the trivial case; for everything else, compare the decoded frames. Upload both files to DiffALL and in under a minute you’ll know whether they match, how closely, and — if they don’t — the exact second things diverge.