Same footage, different file: a duplicate-check test
Two videos can contain the same frames and still be different files. That matters when an upload folder contains original.mp4, final.mp4 and another export that looks identical. We tested two specific cases with an owned sample instead of assuming the filenames told us what was inside.
This is a companion to our short exact-copy demonstration. The original comparison below changes container metadata. The October 7 control adds a separate lossless and lossy re-encode; neither test represents every video editor or export setting.
What we actually tested
1. Same contents, different name
We copied the file and gave the copy a different name.
Result: Same file. Its contents and size stayed the same. The new name did not make a new export.
2. Same decoded video, changed file
We copied the video stream into a new MP4 container and changed its title metadata.
Result: Different files. File contents changed. A separate comparison of the decoded video output matched for this sample.
We ran the comparison function used by our free checker against these real local files. The renamed copy matched; the metadata-changed copy did not. We also compared the decoded video output of the original and the metadata-changed file. We did not measure audio equivalence or test a re-encoded pair in that September experiment. The separate October control below addresses re-encoding for video frames only.
Does a re-encode change the footage? Our 24-frame control
On October 7, 2026, a lossless re-encode produced a different file while all 24 decoded video frames matched. A lossy re-encode produced a different file and zero exact frame matches. Both exports were made from the same controlled two-second source. A different-file result alone did not tell us whether the scene was different.
To make this comparison small and repeatable, we took the first 2 seconds of our owned 8.041667-second sample. We decoded it once, selected 12 frames per second, scaled it to 320 × 568 pixels and stored those 24 frames with the lossless FFV1 codec. That normalized file was the source control for every row below. It was not compared directly with the full-size original: reducing the size and frame rate was already a change.
| Variant | File bytes | Exact frames matched |
|---|---|---|
| Source control | 1,327,785 | 24 of 24 |
| Exact byte copy, different filename label | 1,327,785; same contents | 24 of 24 |
| FFV1 lossless re-encode | 1,327,750; different contents | 24 of 24 |
| H.264 lossy re-encode, CRF 28 | 32,563; different contents | 0 of 24 |
The lossless control used FFV1 level 3; the lossy control used libx264, medium preset and CRF 28. Both kept 320 × 568 pixels, 12 fps and the yuv420p pixel format, and both omitted audio. They were encoded in Matroska containers. The exact-copy control was a byte-for-byte copy held in memory with another filename label, rather than a new media file saved to disk.
We checked full file contents separately from the decoded frames. For the frame check, every input went through the same video-only decoding options. We compared the ordered frame hashes and confirmed 24 frames in each output. The container filenames and title metadata were not evidence of a new scene.
ffmpeg -i source-control.mkv -map 0:v:0 -an \
-pix_fmt yuv420p -c:v rawvideo -f framehash -This is an equivalent command for a saved input; the measurement here passed media buffers through the process without keeping extra exports. FFmpeg version 9.0.1 was used. Its official framehash documentation describes per-packet hashes and the default conversion to raw video. We explicitly requested raw video and the same pixel format so the comparison was about decoded frame data.
What these numbers let you decide
- Same bytes: the exact copy adds no new file contents. Rename it for organization if needed; the label does not add a new edit.
- Different bytes, all frames matching: our lossless export retained the measured video frames. It still needed a separate audio check, which this control did not perform.
- Different bytes and different exact frames: lossy compression changed pixels. Zero exact matches did not mean 24 new scenes or a newly original work.
Use exact-file checks to remove redundant copies from a working list. Use a separate content review when the question is the same scene, the intended edit, captions, sound or publishing rights. Keep the source until that review is complete. These measurements do not judge creative quality, copyright, platform approval or audience response.
Watch the same two-second control, with narration
AI-generated animal footage reused for this controlled test; one sample only, no earnings/quality guarantee.
The experiment selected 24 frames at 12 fps from a two-second control. This narrated explanation is a separate 18.75-second video at 24 fps. It repeats that control to explain the measurements. Zero exact pixel matches after lossy compression does not prove worse picture quality.
The cover is an AI-generated illustration of matching filmstrips in different file containers. It is not a screenshot or measured test frame.
Open the 18.75-second comparison video. Keep your source and compare the intended export before your next upload.
Sources checked
- FFmpeg official framehash documentation, checked October 7, 2026.
- Our own 2-second control derived from an owned sample, measured October 7, 2026 with 24 frames per output. This is a single controlled example, with no audience-performance measurement.
Use the result to answer the right question
- “Did I rename a copy?” An exact-file check helps answer this.
- “Does this export show the same footage?” A different-file result cannot settle that. Review the videos and their export settings.
- “Which version should I upload?” Check the intended edit, language, captions and sound. Matching files alone do not prove that either version is the right one.
Try the check before your next upload
- Keep your source files. Open the free exact-file checker.
- Choose two video files. Each must be no larger than 64 MiB.
- Select Compare files. Read “Same file” or “Different files” before deciding your next step.
- For different files, inspect the actual exports. Do not treat “different” as proof of a different story, edit or quality level.
The comparison code runs in your browser and does not send your selected file contents to a server. There is no tool account to create. Supported extensions are MP4, MOV, WebM, M4V, OGV and OGG.
Watch the renamed-copy example
The earlier 10.5-second demonstration shows the renamed-copy case. It does not show the new metadata test described above.
Open the original demonstration on YouTube
What a hash check cannot tell you
An exact-file check answers one question only: are these the same bytes? Two later measurements on our own files, published in Before You Upload an AI Song or Video, show where that stops being enough. A 540p re-encode of the same footage and a copy with the first 2 seconds cut both have completely different SHA-256 hashes, so a hash check calls them different files. A picture-and-sound fingerprint comparison matched both to the original and placed the trimmed copy at 2.00 s into the source. On our own songs, same-song pairs correlated 0.967–1.000, while different recordings reached 0.393 at most.
Those fingerprint measurements are separate from the free checker on this site, which compares exact file contents only. In practice we use the two results like this:
- Same hash: same bytes. Keep one copy and one filename in the upload note.
- Different hash: it could be a metadata change (test 2 above), a re-export, a trim or genuinely different footage. Open both files, or run a fingerprint comparison, before deciding which one to upload.
One useful folder habit
After checking, record one approved filename in a short upload note: platform, language, approved export, check date. That note gives you a clear next action without relying on a pile of files named “final.” Keep any older exports you still need; the checker does not delete or rename anything.
Tested September 14, 2026 with owned sample files. We built the free checker. This article contains no affiliate product links, and the test is not a customer case or a promise of platform approval.
Comments
Post a Comment