Test a complicated FPV route as a sequence of small spatial decisions before requesting the entire journey. Give each test a clear entrance, one important turn and a readable destination. This helps you find where the scene becomes inconsistent, and whether the idea works better as a continuous shot or a deliberate edit.
Draw a route that can be explained
For a fictional planning example, the viewpoint enters through a cafe door, curves around a table and reaches a person beside a window. Draw the route on a simple plan, then identify the doorway, the near table corner and the final window. Those landmarks are more useful than a decorative spiral with no relationship to the room.
Decide the approximate height and which side of each obstacle the camera passes. A path viewed from above does not specify whether the viewpoint travels under a shelf, over a chair or through it. Treat clearance as an instruction you must explain, not a property that the line automatically contains.
Split at changes in spatial responsibility
| Test | Starting evidence | Main question |
|---|---|---|
| Enter the doorway | Doorframe and interior visible | Does the entrance remain connected? |
| Round the table | Table corner and floor visible | Does the viewpoint pass on the intended side? |
| Reach the window | Person and window visible | Does the finish preserve their relationship? |
Kolbo documents path guidance in its drawing mode. That does not establish collision-safe navigation or a reconstructed room. Likewise, named movements in a camera glossary help communicate intent but cannot prove that an output follows real geometry.
Keep the person's action simple while testing the route. Otherwise a changing pose, a passing customer and the camera turn can all fail together, making diagnosis harder.
Inspect the difficult intervals
A shot can begin and end convincingly while taking an impossible shortcut between them. Watch the doorway crossing and the table's occlusion at normal speed, then step through those intervals. Look for a wall turning transparent, a table corner moving aside, repeated furniture or a sudden change in travel direction.
Write observations with locations: “The near table leg disappears as the viewpoint turns,” rather than “the camera feels wrong.” If the problem starts at the doorway, improve the entrance view or shorten that passage. If it starts behind the table, reconsider the obscured part of the route. Preserve the source image and the rejected clip so you can compare a revision honestly.
Decide between one take and an edited journey
Passing individual tests does not guarantee that a model can connect them seamlessly. If continuity matters, test the complete route after approving the sections. If the story merely needs arrival, an intentional cut can be a better solution than a long unstable flight.
For an edited version, choose a cut where the destination becomes understandable, rather than hiding a broken wall in a random blur. Shared scenery must still agree across the views. Record those relationships in an Imagild film plan; a text plan is not a promise of a built-in drawn-path renderer.
Accept the route when a viewer can describe where the viewpoint went, obstacles remain plausible and the final composition serves the scene. Keep ambitious speed and complex subject action for later iterations. The first useful result is a readable journey, not the longest possible path.