A security architect who flies DJI drones as a hobby has built the flight archive software DJI itself never shipped. DJI Metadata Embedder, a free and open-source Windows app and command-line tool from developer Marcus Westermark, takes a folder of drone footage and produces one interactive map of every flight in it, rendered flat or over real 3D terrain, with the camera’s ground footprint sweeping the landscape as the flight replays.
Westermark tagged DroneXL on X this afternoon with a screenshot of the 3D mode and a four-word question: what to build next? Before answering that, it’s worth explaining what he has already built, because the answer is more than most pilots would expect from a project with 16 stars on GitHub.
The tool reads the SRT telemetry files DJI quietly writes next to every video: GPS position, altitude, speed, and camera settings, roughly 30 data points per second. Then it does something useful with them, which is more than can be said for the software DJI ships.
One Folder In, One Map Out
Point DJI Metadata Embedder at a footage folder and it returns a single interactive map with every flight as its own colored track, complete with start times, durations, and altitude ranges in clickable popups. The flight map feature is still marked experimental while its logic for joining size-split recordings gets tuned on real-world feedback. The same engine pins geotagged photos on a clustered map with thumbnail previews, gives 360° panoramas their own interactive viewer, and writes GPS data directly into video files so Windows Photos and Google Photos can find drone clips by location. No re-encoding, no quality loss.
Exports cover GPX for Google Earth, CSV for spreadsheets, GeoJSON and KML for GIS tools, and Cursor-on-Target for ATAK systems. The parser handles ten fully documented DJI telemetry formats, from the bracketed style of the DJI Mini 4 Pro to the legacy notation of older aircraft, and it already supports the newest hardware: the DJI Mini 5 Pro, Neo 2, and Avata 360, including the sidecar-less models that bury telemetry inside the MP4 itself.
That model list matters because of scale. FAA-funded research published in March found DJI aircraft account for 96% of detected drone platforms in the United States. Every one of those aircraft is generating telemetry its owner has no good first-party tools to use.
Flights Rise From The Terrain As Curtains
The 3D terrain mode, the feature in Westermark’s tweet, draws each flight as a translucent curtain from the ground up to the aircraft, so a 30-meter (98-foot) hover and a 300-meter (984-foot) transit read differently at a glance instead of looking identical on a flat line. A “view from here” function moves the camera to where the drone was, pointed where its gimbal pointed. Press play and it rides the whole flight while the camera’s ground footprint shades the terrain below. Click any spot on the map and it tells you which seconds of recording covered that ground.
For anyone who has ever scrubbed through an hour of footage hunting for the ten seconds that showed a specific rooftop, that last feature alone justifies the download.
A Security Architect Builds Privacy In By Default
Westermark’s day job shapes the tool’s behavior in ways that separate it from typical hobby software. Everything runs locally, with no uploads and no accounts. Releases are code-signed and carry Sigstore attestations proving the binary you download was built from the public source code by GitHub’s own build system, not modified on someone’s laptop.
GPS redaction ships in the box. A single flag coarsens every position to about 100 meters (328 feet) before a map or export is written, and the drone’s HOME point, which is the operator’s launch location and usually their driveway, stays out of every output unless explicitly requested. DroneXL’s recent reporting on the SFPD Skydio feed that streamed on the open internet for six months showed exactly how much a telemetry log reveals: position, altitude, speed, and heading from takeoff to landing. Westermark treats that sensitivity as the default assumption rather than an afterthought.
The Tool Belongs In The Verification Kit
The most consequential feature is the least flashy. A verify-sun command computes the sun’s azimuth and elevation over a clip so an analyst can check whether the shadows in the footage match the astronomical sun for the claimed time and place. Back in 2022, DroneXL published a guide on how to geolocate drone videos to document war crimes using SunCalc and satellite imagery by hand. Westermark has now automated that shadow check into a single command that reads the drone’s own telemetry.
The project’s documentation leans into this. It names open-source verification, conflict documentation, search and rescue, and humanitarian damage assessment as the intended uses, and draws a hard line: “We do not build targeting or other offensive capabilities.” Meanwhile, the commercial world is moving fast to monetize the same data. Cellebrite just locked down an exclusive drone forensics partnership with SkySafe to sell flight-history analysis to more than 7,000 law enforcement and enterprise customers. Westermark gives away the civilian half of that capability under an MIT license.
DroneXL’s Take
DJI records some of the richest telemetry in consumer aviation, then hands its pilots almost nothing to do with it. No flight archive map. No photo map. No verification tools. Ten incompatible telemetry formats across the product line, as if the company never expected anyone to read them. It took one pilot with a security background to build what DJI’s software division has not shipped in a decade of market dominance, and that should embarrass a company whose aircraft make up 96% of the American fleet.
This is also what the drones-for-good argument looks like in practice. We’ve covered lone researchers before, like Konrad Iturbe, whose bot fingerprints DJI shell companies through FCC filings, and the pattern holds: the most interesting accountability tools in this industry keep coming from individuals, not vendors. A free command that cross-checks shadows against the sun is worth more to a conflict researcher in Ukraine than another watermarking press release.
As for Marcus’s question, here’s my answer. Build the compliance layer. Draw the flight track against airspace grids and altitude ceilings so the map doubles as a Part 107 record a pilot can hand to a client or the FAA. Then bring the desktop app to macOS, where half the creators in this industry live. The verification work is the noble half of this project. The logbook nobody has to type into is the half that gets it to a hundred thousand pilots.
Source: CallMarcus.com, GitHub
DroneXL uses automated tools to support research and source retrieval. All reporting and editorial perspectives are by Haye Kesteloo.
