Tiny sensor cameras, focallength below 1mm (unit) , flame cant handle those

Just want to share a “Today I learned” moment.

I used KeenTools PinTool / Geotracker in nuke to track a unknown camera onto some geo, all keyframed by hand but this would apply to regular Geotracker results …

KeenTools is pretty great at aligning geometry and guessing the focal length, but without any known filmback, it just goes oh yea the filmback is 1 unit… and continunes to only use focallength to calculate the Angle of view

What happens is geotracker writes a camera out with a sensor size of 1x1 unit (mm in nuke and flame) and a focallength of like 0.3 or something, now every program I tried is fine with this, as its a valid thing that checks out. Nuke, 3DSmax, Blender…

All but Flame. Flame decides to completely murder the scale of the camera, i think it cant deal with focallength below 1 even though the AOV checks out based on the small filmback.

so just another thing to check, sometimes its not just scene scale but “clipping” values in the focallength or filmback that causes cameras to not align.

On another note , its called FOV in flame, Field of View depends on the distance though , shouldnt it be correctly named angle of view?

Also while at it: Filmback should be able to be locked to action background ratio :slight_smile: just for the beauty of it… a 1x1 filmback with action size of 1920x1080 makes no sense to me. (same complaint with nuke, 3DE can do this the best)


If you don’t know the camera (or know just a sensor size), you may still always connect the default Nuke’s camera (22.3x17.4 or with adjusted sensor size) and focal length estimation will use this sensor size and not default 1x1mm


thats good to know! I thought it might go crazy if I do this with focal length estimation on.

I never had a problem either way , but still would be a good idea to default to like 30x30 or something in geotracker, its not a big deal either way :wink: