Fog

Fog is a 2D Deep Pixel post-processing node for rendered images that contain a Z-buffer. It colors atmospheric haze according to depth, not a volumetric object inside a classic Fusion scene3d graph.

The Resolve 21 reference specifies three inputs: required orange Input (image with Z), optional green Blur Image (2D source of fog variation), and optional blue Effect Mask (mask data). The original site article listed only two ports and therefore omitted a major feature.

A useful Z channel normally comes from Renderer3D with Z Depth output enabled or from an EXR whose depth channels were correctly mapped into the Fusion auxiliary Z buffer. A valid 2D wire alone does not prove that the required per-pixel depth data exists.

Z Near Plane marks where fog thins out to nothing; Z Far Plane marks the distance where fog contribution becomes opaque. The archived Inspector shows Near 0.0 and Far -500.0. These are example values, not universal meter readings or a promise that every external EXR uses negative far depth.

The Sample buttons under Near/Far take numeric Z values from points in the Viewer. Pick stable geometry, not transparent borders or empty backdrops. The Fog Color eyedropper selects color instead; it is not a depth sampler.

Z Scale expands or contracts values in the depth channel, allowing adjustment to a different upstream scale. It cannot invent a missing Z pass or fix an incorrectly imported depth channel.

With no green Blur Image connected, Fog is a uniform base tint modulated by depth. Attach FastNoise or another 2D image to Blur Image to supply spatially varying fog pattern. This auxiliary image is not a 3D smoke density field.

Fog Color chooses the RGBA haze tint. Red, Green and Blue set the color balance, Alpha changes transparency of the chosen fog color, and Fog Opacity scales the overall fog contribution across channels. Evaluate alpha and opacity separately in a transparent image pipeline.

The optional blue Effect Mask limits the processed image region after the fog operation. Rectangle.Mask must connect to Fog.EffectMask, not to the green Blur Image port, which expects image data.

After Fog, a ColorCorrector and MediaOut can finish the composite. To keep titles crisp and unaffected by scene depth, place Text+ into a downstream Merge.Foreground instead of running pre-composited UI through the depth effect.

Fog is not Fog3D. Fog3D has a required scene3d input and outputs scene3d to Renderer3D; ordinary Fog receives and outputs a 2D image after the render boundary. Connecting 3D scene output directly to Fog is a type error.

Typed tutorial graphs can check the declared image/mask/scene3d connectors and topology, but not verify that Z exists in the pixels, EXR mapping, alpha math or actual rendered appearance. Those checks remain pending in native DaVinci Resolve.