"Scanning 20,000,000,000+ video nodes..."
A browser-based tool for surfacing forgotten, unlisted, and unindexed YouTube videos by generating raw camera filename queries — the kind of titles people never change before uploading.
Built on the observation that countless personal videos were uploaded directly from phones and cameras with default filenames like DSC_0247, IMG_3891, GOPR0012, or Desktop 2014 03 22 — and are effectively invisible to normal search.
Most people who upload home videos, travel clips, and personal recordings don't rename their files. YouTube indexes these anyway. By searching for the exact filename patterns cameras generate, you can find videos with zero SEO optimization — raw, unseen, and often years old.
my tool automates this process by randomly generating filename-pattern queries and firing them at YouTube search across different temporal layers.
| Layer | Description | Date Filter |
|---|---|---|
| Layer I — Recent Fallen | Videos uploaded in the last few years, already forgotten | before:2023 |
| Layer II — Forgotten Era | Mid-2010s uploads, peak unoptimized uploading era | before:2016 |
| Layer III — Pre-Archive | Early YouTube era, pre-smartphone dominance | before:2010 |
| Pseudo Random Node | No date filter — pure random across all time | none |
Generates queries based on real camera/device filename conventions:
IMG XXXX/DSC XXXX/DSCF XXXX— standard camera namingGOPRXXXX/CIMGXXXX— GoPro and Casio formatsMVI XXXX— Canon video formatVID YYYYMMDD/Desktop YYYY MM DD— date-stamped exportsGH XXXXXX— GoPro Hero formatclip XXXX/video XXXX/File XXXX/export YYYYMMDD
Fine-grained query control:
- Prefix Term — lock to a specific filename prefix from the pool
- Suffix / Keyword — append a tag (e.g.
birthday,vacation,vlog) to narrow context - Token Format — choose between 4-digit, 6-digit, date (
YYYYMMDD), spaced date, or none - Date Filter — set
before:,after:, or abetweenrange manually - Sort Mode — sort by upload date, view count, rating, or relevance
- Mutation — randomly alters characters on each rescan for variation
- Open ×3 At Once — fires three unique queries simultaneously
- Live Query Preview — see the exact search string before firing
- Force Corruption — visual glitch effect
- Blackout Layer — unlocks after 8+ scans, switches to unsorted archive access mode
- Copy Query — copies the current advanced query to clipboard
No installation required. Just open index.html or use the github pages link in any modern browser.
1. Open terminal.html / github pages site
2. Select an archive layer
3. Click SCAN NODE
4. Browse what surfaces
For more targeted digs, open the [ ADVANCED ] panel and configure your query before firing.
The tool wraps queries in quotation marks to signal exact-match intent. However, YouTube does not reliably honor quote operators — its search engine prioritizes engagement signals over textual precision. If you're getting irrelevant results, this is a platform limitation, not a bug.
Workarounds:
- Use Google video search (
site:youtube.com "DSC 0247") — Google's index of YouTube actually respects quotes - Add negative keywords in the suffix field (e.g.
-tutorial -review) - Use the Upload Date sort mode to surface recent raw uploads first
Core concept inspired by KVN AUST and the broader practice of "lost media" YouTube archaeology — the idea that the platform's index contains millions of videos that exist in a kind of digital limbo, uploaded and immediately forgotten.
Do whatever you want with it. No tracking, no backend, no dependencies — just a single HTML file. Have fun, Explorers! :)