gmatocha/dji_04_air_unit_gyro_patcher — explained in plain English
Analysis updated 2026-05-18
Repair shaky DJI Air Unit footage caused by corrupted gyro telemetry bursts.
Extract raw gyro data from a DJI MP4 for inspection or custom processing.
Feed cleaned telemetry into Gyroflow or similar stabilization software for smoother results.
Detect suspicious gyro spikes and review them before deciding whether to patch.
| gmatocha/dji_04_air_unit_gyro_patcher | 0-bingwu-0/live-interpreter | 0cm-labs/tokenizer-benchmark | |
|---|---|---|---|
| Stars | 2 | 2 | 2 |
| Language | Python | Python | Python |
| Setup difficulty | moderate | moderate | moderate |
| Complexity | 3/5 | 2/5 | 2/5 |
| Audience | vibe coder | general | researcher |
Figures from each repo's GitHub metadata at analysis time.
Requires FFmpeg and Python 3.8+, plus footage originally stabilized in Gyroflow.
DJI_04_Air_Unit_Gyro_Patcher is a set of Python scripts that fixes a specific stabilization bug in video files recorded by DJI Air Units, mostly ones using the 1469D gyro chip found in 2026 hardware. Some recordings from these units contain short bursts of bad gyro telemetry that cause visible shakiness even after stabilization software like Gyroflow has processed the footage. This tool detects those bad stretches of data and repairs them directly inside the MP4 file. The main script, dji_gyro_fix.py, runs the whole process in one step: it extracts the gyro telemetry from an input MP4, cleans up the bad regions, and writes the corrected data back into a new patched copy of the video. Separate scripts are also provided for extracting telemetry to a CSV file, patching a CSV on its own, and injecting a patched CSV back into a video, for users who want to work through the steps individually. Detection works by measuring the gyro's overall movement magnitude for every sample and comparing it against each file's own normal baseline, rather than a single fixed threshold. Regions where the magnitude spikes far above baseline are flagged as glitches, then the flagged area is grown outward to catch the fading wobble that often lingers after the main spike, stopping once the data has been calm for a set amount of time. A safety check treats unusually long high magnitude regions as possible real camera motion instead of a glitch, and reports them separately rather than silently altering them. Once a bad region is identified, the tool bridges across it using a smooth interpolation between the last good reading before the glitch and the first good reading after it, then recalculates the rest of the orientation data for the whole file. It does not attempt to guess what actually happened during the glitch, it simply smooths over it so stabilization software has clean data to work with. The tool requires FFmpeg and Python 3.8 or newer, and is meant for drone pilots and video editors dealing with shaky DJI Air Unit footage. The author notes that the code and documentation were generated with Claude Sonnet 5 and may contain errors.
Python scripts that detect and repair short bursts of bad gyro data in DJI Air Unit video files, fixing shakiness that survives normal stabilization.
Mainly Python. The stack also includes Python, FFmpeg.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly vibe coder.
This repo across BitVibe Labs
double-check against the repo, no cap.