Displace3D

What Displace3D does. The node converts a height map or a vector field into actual changes in mesh vertices. Unlike the 2D Displace node, it changes the silhouette, object intersections and spatial relationships before Renderer3D.

The inputs are not interchangeable. Required orange SceneInput receives a 3D object or scene; green Input receives an optional 2D displacement image. With no map image, the tool effectively passes the scene through unchanged.

Typical setup. Subdivided Shape3D → Displace3D.SceneInput; FastNoise → Displace3D.Input; Displace3D.Output → Merge3D with Camera3D and lighting → Renderer3D → MediaOut. A camera can be merged into the scene upstream when Point to Camera is needed.

Mesh density is the hard limit. Displace3D moves existing vertices and performs no tessellation. Even a 4K height map cannot create fine relief on a mesh made of only a few polygons; add subdivisions to the source geometry upstream.

Where map values are sampled. Every vertex uses its UV/texture coordinates to look up the 2D image. If the pattern is shifted or stretched, inspect source UVs, UVMap3D projection and the map first; large Scale values will not fix wrong coordinates.

Channel: scalar modes. Luma, Red, Grn, Blu and Alph extract scalar magnitudes from the selected image channel. With Point to Camera off, vertices move along their normals, so reversed or inconsistent normals can reverse the apparent deformation.

Channel: Z and Pos. These modes rely on genuine auxiliary depth or position data being present in the image stream. A grayscale RGB preview does not automatically become AUX Z, and choosing Pos does not manufacture a World Position pass.

Channel: RGB. This is vector displacement: the three color components represent XYZ displacement components rather than a single averaged brightness. Validate axis order, sign and scale on simple test geometry before accepting a map from another 3D package.

Bias and Scale are sequential. Bias first adjusts the displacement baseline, and Scale then multiplies that result. Remember Bias → Scale: high Scale magnifies both useful relief and an incorrect neutral point.

The map neutral point is source-dependent. For some normalized 0–1 height maps, Bias near −0.5 is a useful diagnostic starting point, not a universal value. Floating-point maps can also contain valid negative displacement values.

Scale is the amplitude control. Start small and compare both the camera view and a free 3D side view. Increasing Scale cannot compensate for an undersampled mesh and may magnify seam cracks.

Point to Camera off. Conventional scalar displacement follows each vertex normal. This suits terrain, waves and organic surfaces; the direction of normals on the underlying mesh still matters.

Point to Camera on. Each vertex moves toward the selected camera instead of along its normal. This can deform an image plane into depth while preserving its appearance through the chosen camera, enabling 3D layering and occlusion.

Camera selection. The Camera menu selects a camera already in the scene connected to SceneInput; Displace3D has no separate camera socket. Merge Camera3D into that scene upstream when using Point to Camera.

Two complementary views. Inspect through the selected Camera for compositing, and use a free Perspective/Wireframe view for actual shape. A final beauty render alone can hide incorrect displacement direction.

Mesh density and normals are different problems. A coarse silhouette generally indicates insufficient subdivisions; wrong deformation direction on a dense mesh suggests normal orientation or map values. Inspect the source geometry and use ReplaceNormals3D selectively.

Normals after deformation. Strongly displaced surfaces can shade poorly even when the shape is correct. Assess recomputed or replaced normals after Displace3D separately from the original normals that set the scalar displacement direction.

Seam cracks. Nearly coincident, unwelded vertices with different normals may pull apart under displacement. Weld3D upstream can help after that defect is confirmed; Weld3D is not a polygon reducer and should not be inserted blindly.

Particle caveat. Passing particles through Displace3D disables Always Face Camera, because each particle consists of four vertices that are individually displaced. This is documented behavior rather than an arbitrary rendering failure.

Node order. UVMap3D before Displace3D affects map sampling, ReplaceNormals3D before it affects the original scalar direction, and Transform3D before versus after changes scene placement. Adjust one factor at a time when troubleshooting.

Animated map. Evolving Fast Noise or an animated source image deforms the geometry over time. Verify stable UV mapping, object intersections and possible popping during playback, not on a single still frame.

The 3D-to-2D boundary. Displace3D Output stays scene3d. Renderer3D must convert it to an image for MediaOut; alpha, lighting and rendering behavior depend on the overall render setup.

Performance. The processing cost is strongly affected by vertex count and scene complexity, not only map resolution. Choose density for the actual shot and test several representative frames.

Validation boundary. Channel and Camera controls are supported by the captured Inspector and documentation. Exact external RGB/Pos/Z pass conventions, camera corner cases and imported meshes still require testing in the target Resolve build; text audits are not substitutes for a live render.