Skip to content

Insights · Shot planning

AI plans the shot. And the route behind it.

AI illustration: a drone and controller on a planning desk; a hand reviews a route alongside light, weather and camera symbols.
AI illustration
Published

In short · 4 min read

  • A developed toolchain from a visual brief to a route file.
  • AI plans, software calculates, and a person reviews and supervises the flight.
  • A documented still/video test, with device-specific checks.

The shot starts before take-off

A camera drone is capable of a lot. A good shot still needs a plan: where is the sun, what should stay in frame, and how should the camera and aircraft move together?

We developed an AI-assisted toolchain for that task. AI turns the visual brief into a shot list, chooses movements and prepares settings. A program calculates waypoints and camera actions. A person reviews the plan and supervises the flight.

From the intended image to camera movement

“Show a building, then open the view towards the water” is a visual brief. Movement, camera direction and timing turn it into a planned shot.

Visual idea What we plan
Show a subject in context An arc around it, viewing direction and camera tilt
Reveal the surroundings A movement that opens the view gradually
Create a calm overview Path, altitude and speed suited to the framing
Combine stills and video Capture actions at suitable waypoints
AI illustration of a fictional coast with schematic flight paths and viewing directions.
AI illustration: a shot idea and camera movement at a fictional location. Not an approved flight route.

What AI decides and what the program calculates

AI does more than draw an attractive line on a map. It works from the purpose, subject, light and device capabilities. Our developed workflow connects several steps:

  • Assess conditions: Consider solar position, weather data and available terrain information.
  • Design the shots: Specify camera movement, orientation, tilt and still or video actions.
  • Calculate the route: Turn chosen parameters into waypoints and a route file for the tested device setup.
  • Check the plan: Check selected configured limits and display warnings.
  • Review before take-off: Inspect the route and viewing directions in a preview and check them on the device.

The generated plan can include zoom values too. Whether every movement executes as intended on a particular device is a separate test. Support in software and a successful camera test are different forms of evidence.

AI illustration: a pilot checks a route on a controller at a fictional coast while the drone rests with stationary propellers on a case.
AI illustration: before take-off, the person checks the route, equipment and surroundings. The coast and pilot are fictional.

Built, checked and flown

The workflow was developed for a DJI device setup. A documented test flight on 11 November 2026 successfully switched between still capture and video. During our code review on 19 November 2026, the 16 existing tests all passed.

Those tests cover geometry, solar calculations, selected planning constraints and the route-file format. That is concrete technical evidence. It does not establish that every movement, zoom transition or new device combination has been tried.

The coastline shown here is an illustration of a fictional location. It explains planning; it is neither footage from the test flight nor a route to fly.

Supervision is part of the system

The person remains the pilot. They assess the situation on site, check the route and equipment, and can intervene. A planning check does not automatically detect every obstacle or change in the surroundings. Weather and terrain data cannot replace an on-site assessment.

Transfer also depends on the controller and software version. Our developed workflow is therefore not a blanket promise for every drone model or a complete clearance to fly.

What carries over to other tasks

The example demonstrates a capability beyond drones: AI can turn a creative task into checkable steps and executable device data. We connect a language model, calculations, interfaces and human judgement to do that.

For a new task, we first establish the required shot and available equipment. Then we develop an appropriate solution, separating planned functions from results that have actually been tested.

Tell us about your task or learn to use AI on your own project. Our Project Bantayan insight offers another example of custom software development.

Sources.

  1. SolvraONE: entwickelte Drohnenplanung, Code-Abgleich und TestsOwn source · As of
  2. Entwicklung und dokumentierter TestflugOwn source · As of