Camera3D

Camera3D has three optional inputs with different data domains. The orange SceneInput accepts an existing 3D scene or object. The magenta ImageInput accepts a normal 2D image and exposes the conditional Image, Materials and Projection tabs. The green RightStereoCamera input accepts another Camera3D for an external right-eye camera in a stereo rig.

The Controls tab separates two fundamentally different projection models. Perspective behaves like a conventional camera: Focal Length, Film Back and Angle of View are linked, and object distance affects apparent size. Orthographic removes perspective diminution; Viewing Volume Size controls the framed volume, so moving the camera along Z does not by itself make objects larger or smaller.

Near and Far Clip define the visible depth range. In Perspective, their manual values are used when Adaptive Near/Far Clip is disabled; keeping the useful range reasonably tight improves depth precision. If geometry suddenly disappears or distant depth artifacts appear, clipping should be checked before blaming materials or Renderer3D.

Film Back describes the virtual sensor geometry through Film Gate, Aperture Width/Height, Lens Shift X/Y and Resolution Gate Fit. Focal Length alone is not enough to reproduce a field of view because the aperture size is part of the optical relationship. Camera matching, imported cameras and overscan therefore require focal length, film back and gate fitting to remain consistent.

Connecting ImageInput enables two different image workflows. Image Plane creates real camera-attached geometry aligned to the camera field of view; it is not merely a Viewer reference overlay. It exposes the common 3D object controls for visibility, lighting/shadows, matte, blend, normals/tangents and ObjectID, while the separate Materials tab exposes the full common material block. Camera Projection instead projects the incoming image through the camera onto 3D geometry. Light and Ambient Light projection require lighting to be evaluated in Renderer3D, while Texture mode is intended for a relightable projection workflow with a compatible Catcher/material setup.

The Stereo block builds left/right views inside Camera3D. Toe In rotates the eyes toward a convergence point, Off Axis offsets the frusta without the same vertical-parallax problem, and Parallel keeps the cameras parallel and does not use Convergence Distance. RightStereoCamera can replace the internally generated right eye with a separate Camera3D.

Transform moves the virtual camera as an object inside the 3D scene. Control Visibility exposes frustum, view vector, clip, focus and convergence guides for staging and diagnostics only. To inspect the actual shot through Camera3D, normally view a downstream Merge3D or Renderer3D and select the intended camera; viewing the Camera3D node itself without useful scene content is not the same as evaluating the rendered shot.