Skip to content

Calibration

The Calibrate tab recovers a camera's lens and orientation from straight lines in the picture, the way fSpy does it. It works on a single still image, or on a video session to feed a better FOV into the solve. If you've ever lined up perspective guides over a photo to matchmove a matte painting, this is that, automated.

Why it works

Parallel edges in the real world (the sides of a doorway, floor tiles, a tabletop) converge in the image toward vanishing points. Tell Kitly which image edges are parallel in reality, and it can compute the field of view, the focal length, and the camera's orientation from the geometry alone.

Calibrate an image

  1. Open the Calibrate tab (in a still-image session, or a Camera Scene session).
  2. Drag the first pair of lines along two real-world-parallel edges. This pair is the X axis, drawn red. Pick long, clearly straight edges; window frames and skirting boards beat curtains.
  3. Drag the second pair along edges running the other horizontal direction. This is the Z axis, drawn blue.
  4. Optionally place a third pair for the Y axis (vertical edges), drawn green, for extra accuracy.
  5. Watch the live readout update as you adjust: FOV, focal length, and camera height.
  6. If the world comes out mirrored or turned, use the axis flip toggles.
  7. Click Save Calibration.

Metric scale

By default the scene has no real-world size. To anchor it:

  1. Pick an axis for the reference measurement.
  2. Enter the real-world length of that stretch (a door is about 2 meters, a countertop about 0.9).

Now the calibration carries metric scale. For camera solves, this is also what turns the solve's arbitrary scale into real units, so a blockout export drops into Blender at the right size.

Still-image cameras

A still-image calibration session ("New calibration from image") produces a static solved camera. Its tabs are Calibrate and 3D. After calibrating, use the 3D tab to place blockout geometry against the photo, exactly as in a camera session's blockout step, and export the camera and meshes as a GLB.

This is the fast path for environment work from a single photo: calibrate the shot, block the set, export, and model over the blockout in your DCC with the photo matched behind a correct camera.

Feeding a camera solve

In a video session, calibrating gives you a measured FOV instead of a guessed one. Click Apply FOV to Solve and the tracker uses the calibrated FOV as a known constant, which stabilizes the solve the same way a manual lens entry does. Calibrate on a frame where straight architecture is clearly visible.