Bilawal Sidhu reconstructs Fleet Week in a 3D digital twin using flight and vessel telemetry

• Fleet Week became a telemetry-driven 3D film: the camera follows jets and times speed ramps, according to the demo announcement.
• Blue Angels coverage has a tracking gap: Sidhu says in a reply that the team does not fly with ADS-B tracking.
• God’s Eye View has public code and a hosted version still to come, as Sidhu explained when sharing the project links.

Sidhu says the complicated camera rigs he once made in Maya can now be conjured up with code. He also describes embedding YouTube and X clips in geographic context, without downloading them first.

Bilawal Sidhu rebuilt Fleet Week from real flight and vessel tracks inside a photorealistic San Francisco digital twin.

Fleet Week reconstructed above San Francisco Bay

His camera engine reads aircraft telemetry to track jets and time speed ramps, a lovely piece of automated cinematography.

For live aircraft, God’s Eye View uses another smoothing technique: it renders flights one polling interval behind real time so it can interpolate between updates, according to the project README.

Sidhu described two experiments for subsequent renders:

• Viewpoint: Plane-mounted cameras for Top Gun-style shots.
• Lighting: Sun-angle adjustments across the water and wings.

The Blue Angels don’t fly with ADS-B tracking, Sidhu said when asked about the team.

He did not describe a replacement data source for their flight paths. The thread leaves how those missing tracks are handled in the reconstruction unexplained.

Community contributors have extended God’s Eye View with:

• Hurricane tracking and uncertainty cones.
• Satellite imagery comparisons across dates.
• Local aircraft tracking through a software-defined radio (SDR) antenna.

Director Mode brings geolocated video, satellite imagery and 3D reconstruction into a storytelling workspace. Sidhu describes an agent-assisted authoring workflow in his full walkthrough:

1. Scene authoring with a coding agent, using the scene editor’s reference implementation.
2. Publishing the scene.
3. Screen-recording the result.

The code is available in the GitHub project. The app supports installation through Pinokio or a terminal setup, with optional API keys added inside the app.

Sidhu is taking hosted and pro sign-ups through the Halfpixel waitlist, as he explained in his follow-up.

MIT licensing covers the source code only; third-party data and visual assets retain their own licenses and terms, according to the data-source guide.