Parse DJI drone .SRT subtitle telemetry into GPX / KML flight tracks for Google Earth, or burn an altitude + speed HUD into the sibling video with ffmpeg.

Most DJI models (when Video Subtitles / Captions are enabled) write a .SRT sidecar next to each clip. This tool reads those subtitles only — it is not a flight controller or airdata decoder.

Author: Zach Telford (ztel42)

1. Parse real DJI .SRT dialects and extract lat/lon/altitude (and speed when present).
2. Export GPX 1.1 tracks and KML 2.2 LineString + gx:Track for Google Earth.
3. Burn a simple ASS/SSA HUD (altitude, speed, optional lat/lon) into video via ffmpeg -vf ass=....

Source video and SRT files are opened read-only. Only the output paths you name are written.

Optional recorded speed tags also accepted when present: H.S …m/s, [h_spd: …], [speed: …].

Tolerant of missing optional fields. Cues with no lat/lon are skipped. Altitude prefers absolute (abs_alt / GPS tuple) when present, else relative (rel_alt). Speed uses a recorded horizontal value when the SRT has one; otherwise ground speed is estimated from successive GPS points and timestamps (haversine) and labelled as estimated.

• Python 3.10+ (stdlib only for parse / GPX / KML)
• ffmpeg (+ ffprobe) for the burn HUD command only — install via your OS package manager (apt, brew, dnf, …). The tool does not auto-install packages.
• pytest for tests

cd FlightTelemetryTools
pip install -e .
# or run without install:
PYTHONPATH=. python -m flight_telemetry --help

Entry points after install: flight-telemetry and python -m flight_telemetry.

# Export both GPX and KML
flight-telemetry parse flight.SRT --gpx out.gpx --kml out.kml

# Same via module
python -m flight_telemetry parse flight.SRT --gpx out.gpx --kml out.kml

# Burn altitude/speed HUD (uses sibling VIDEO.SRT if --srt omitted)
flight-telemetry burn video.mp4 --srt video.SRT --out video_hud.mp4

# HUD without lat/lon line
flight-telemetry burn video.mp4 --srt video.SRT --out video_hud.mp4 --no-latlon

Bad input (missing file, no GPS frames, ffmpeg unavailable for burn) exits non-zero with a clear stderr message.

• GPX: one trk / trkseg of trkpt elements with lat, lon, ele, and time.
• KML: a LineString path plus a gx:Track (when + gx:coord); altitudeMode is absolute or relativeToGround matching the altitude source.
• HUD video: ASS overlay in the upper-left showing alt: … m (abs|rel), spd: … m/s (rec|est), and optionally lat/lon, timed to SRT cues.

pip install -r requirements.txt
python -m pytest

Fixtures under tests/fixtures/ cover classic font-wrapped, bracket Mini-style, and GPS-function / H.S dialects, plus missing-GPS cues. The burn test generates a short silent lavfi MP4 (no drone footage required) and skips if ffmpeg/ffprobe are not already on PATH.

FlightTelemetryTools/
  flight_telemetry/
    parser.py       # DJI SRT dialects
    speed.py        # haversine estimate
    export_gpx.py
    export_kml.py
    hud.py          # ASS + ffmpeg burn
    cli.py
  tests/
    fixtures/
  README.md
  pyproject.toml
  requirements.txt

• Not a flight controller. Telemetry comes from video subtitle sidecars only.
• GPS lock gaps mean skipped frames; indoor / pre-lock cues often have no coordinates.
• Estimated speed is approximate (GPS noise, irregular cue timing). Prefer recorded H.S / [h_spd] when the dialect provides it.
• HUD burn re-encodes video through ffmpeg; audio is copied when present.
• Does not decrypt DJI flight logs (.txt / DAT) — only .SRT subtitles.

• Sun Nov 1, 2026 ET — Raised the test dependency floor to pytest>=9.0.3 in requirements.txt and pyproject.toml (older pytest has a known flaw). Added to Dependabot (weekly pip). Tests pass (17).

• Wed Oct 7, 2026 ET — Removed silent apt-get auto-install of ffmpeg. HUD burn now errors clearly if ffmpeg/ffprobe are missing; install them yourself via your package manager.

For portfolio demonstration. Use only with footage and sidecars you own or are authorized to process.