geolocate-from-pixels
Geolocate and chronolocate a photo or video from visual evidence alone — plate and phone number formats, road markings, utility poles, bollards, signage typefaces, architecture and vegetation for place; shadow direction and length with SunCalc for time and date. Use when asked where or when a picture was taken, to verify a claimed location without GPS or EXIF, or to match a scene against Google Earth, Street View, Yandex Panoramas, Mapillary or KartaView. Applies to GEOINT and conflict monitorin
Security Assessment
Detected risks:
About geolocate-from-pixels
geolocate-from-pixels is an OSINT analysis skill for placing where — and when — a photo or video was taken using only its visual content, with no GPS or EXIF required. It counters the beginner mistake of searching before inventorying: the method is to extract every clue first, rank clues by how much of the planet each eliminates, and only then search. A tiered clue table runs from high-value readable proper nouns, phone-number formats, language and script down through license-plate formats, driving side, road markings, utility poles, signage typefaces, architecture, and vegetation.
The documented procedure inventories clues (including informative negatives), fixes the country by combining tier-1/2 evidence, reads on-image text properly (transcribe, translate, then search verbatim in the local language), narrows to a locality, and — crucially — queries map geometry directly with OpenStreetMap/Overpass for structures like a water tower beside a rail crossing rather than panning aimlessly. Candidates are confirmed against satellite and street-level imagery on invariants (building footprint, roof color, window spacing) and chronolocated via sun position and season, with a discipline of requiring three independent aligning features before accepting a match. Reference files cover chronolocation and regional indicators.
It targets GEOINT analysts, conflict monitors, journalist fact-checkers, insurance/claims verifiers, and evidence reviewers, and names Google Earth, Street View, Yandex Panoramas, Mapillary, KartaView, and Bing as imagery sources. The skill is analytical and read-only over public imagery, but geolocation of imagery is dual-use and can help de-anonymize or locate people from their photos, so it carries moderate privacy sensitivity.
FAQ
How does the method work?
Inventory every visual clue before searching, rank clues by how much of the world each eliminates, fix the country from the strongest tells, read and translate on-image text, narrow to a locality, and confirm against satellite and street-level imagery on features that don't change.
What tools and sources does it use?
It uses OpenStreetMap/Overpass geometry queries and imagery from Google Earth and Street View, Yandex Maps and Panoramas, Mapillary, KartaView, and Bing, plus SunCalc-style sun-position reasoning for time and date.
What are its limits?
It is a manual analytic methodology, not an automatic geolocator; it depends on the analyst's reference knowledge and patience, contradictory clues must be reasoned about, and confidence should rest on at least three independent aligning features.
Is there a privacy consideration?
Yes — inferring where and when imagery was captured can help locate or de-anonymize the people or places in it, so it is a dual-use capability best applied within a documented, lawful objective.
Install geolocate-from-pixels
Quick Setup:
- Copy the skill folder to
.claude/skills/ - Claude will automatically detect and use the skill
Repository
useosint/osint-skills