Overlap · how to use
Three steps. No pixels leave your machine.
Fingerprint a local video folder, check it against hashes you already own, then screen the same file against open robotics data.
Once
Install Overlap
Needs Python 3.10+. On Mac, skip system pip3 if it is still 3.9 — use uv.
terminal
brew install uv
uv tool install git+https://github.com/World-Archive/overlap
export PATH="$HOME/.local/bin:$PATH"
overlap --versionYellow AVFFrameReceiver warnings on Mac are harmless. Ignore them.
01 · Generate hashes
Hash a local video folder
Point Overlap at the folder that holds the new footage. It writes a compact .ovlm fingerprint file — typically kilobytes per hour, not the videos themselves.
Input
/path/to/local/videos
Your MP4, MOV, MKV, or MCAP files.
Output
newhash/newhash.ovlm
Share this file. Not the videos.
generate .ovlm
VIDEO_FOLDER="/path/to/local/videos"
NEW_HASH_FOLDER="/path/to/newhash"
rm -rf "$NEW_HASH_FOLDER"
mkdir -p "$NEW_HASH_FOLDER"
overlap --index "$NEW_HASH_FOLDER" index "$VIDEO_FOLDER"
overlap --index "$NEW_HASH_FOLDER" export \
-o "$NEW_HASH_FOLDER/newhash.ovlm" \
--anonymize-paths --stride 1Argument order matters: --index <hash folder> index <video folder>. If you swap them, Overlap indexes 0 files.
02 · Compare against master
Test the new hash against your inventory
You keep a folder of existing inventory hashes — as many .ovlm files as you have. Overlap imports every one of them, then compares the new offer against the full set.
masterhash/
inventory-1.ovlm
inventory-2.ovlm
…
newhash/
newhash.ovlm
compare vs masterhash/
MASTER_HASH_FOLDER="/path/to/masterhash"
NEW_HASH_FOLDER="/path/to/newhash"
WORK_INDEX="/tmp/overlap-master-index"
rm -rf "$WORK_INDEX"
mkdir -p "$WORK_INDEX"
overlap --index "$WORK_INDEX" import \
"$MASTER_HASH_FOLDER"/*.ovlm --label "master inventory"
overlap --index "$WORK_INDEX" compare \
"$NEW_HASH_FOLDER/newhash.ovlm" \
-o duplicate-report.json \
--html duplicate-report.html \
--openOpens duplicate-report.html. If a seller already sent you an .ovlm, skip step 01 and use that file in the compare command.
| Report | Meaning |
|---|---|
| 0% overlap | Nothing matches your inventory — likely original to you. |
| Partial match | Some offered footage is already in your inventory. Price the remainder. |
| 100% overlap | The entire offer is duplicate. Walk away. |
03 · Compare against open data
Screen the same hash against public catalogs
Step 02 is private — your inventory only. This step asks a different question: is this footage already in open robotics datasets?
- 01Open the Overlap catalog.
- 02Drop newhash.ovlm on the page.
- 03Start the comparison.
- 04Read overlap % and matched public datasets.
Website comparison checks public catalog data. The local command in step 02 checks your private master inventory. Run both for full diligence.