Rasterizing
Bake shapes, geometry, a field or a volume into a flat image that lives in the scene, ready to feed feedback loops, filters and content-space transforms.
The Rasterize node takes whatever is wired in, Analytic shapes, 3D geometry, splats, a colour field or a volume, and renders it into a raster image. That image becomes ordinary data in the scene, so it can be blurred, transformed, or looped back into itself.
Why bake to a raster
Feedback and ping-pong trails. Wire
circle → Rasterize → feedback → back in, and each cook stamps the previous frame, leaving glowing trails. See Loops & feedback.Content-space transforms. Once a shape is a raster you can scale, rotate and slide the pixels rather than the shape's control points, for warps and smears.
Filters and post. Image filters expect pixels. Rasterize is how vector shapes and 3D scenes get there.
Caching a look. Bake an expensive build once and reuse the flat result downstream.
The baked image lives in the scene exactly as a loaded image does. See Rasters (images).
2D and 3D content
Flat 2D content, meaning Analytic shapes, text and drawn paths, bakes to fit the canvas with no camera needed. 3D content is captured through a camera: wire a Viewport Camera into the camera input to choose the viewpoint, or leave it unconnected and the content is auto-framed.
Feeding a 3D mesh straight in keeps the conversion on the GPU, so a sphere or an imported model bakes cleanly. If a 3D object vanishes on bake, check that it reaches Rasterize as real geometry rather than already flattened to loose points upstream.
Slicing a field or a volume
A field or volume has no flat image of its own, so wire a Transform into the plane input and drag it to slice one out. projection decides what the slab contributes: First takes the flat slice, Max gives a maximum-intensity projection, and Min, Average and Sum do what they say. The slab's thickness is the plane Transform's own scale along its normal, so scale the gizmo to thicken it, and samples is how many times that thickness is read: 1 for a flat slice, more for a smoother projection. Both controls appear only once a plane is wired.
Pixel format and size
resolution_mode is Canvas by default, so the bake matches the canvas. Custom takes res_width and res_height, Scale a multiple of the canvas, and Fit Content sizes to what is fed in.
format matters more than it looks. It is 16-bit float, and it needs to stay float for the feedback and accumulation loops above: an 8-bit bake clips at white and quantises every pass. supersample renders at 2× or 4× and downsamples for clean edges. background is Transparent unless you set it to Solid.
Raster or sprite
The mode control decides what comes out.
Raster (default) gives the baked image as data, the path for feedback, filters and content-space transforms described above.
Sprite places the baked image into the 3D scene as a textured point. With
face_cameraon it is a camera-facing billboard: cheap, flat, and it scales to millions. With it off it is a fixed quad you can tilt with point orientation.
Rasterize bakes exactly one sprite. To scatter many copies, feed it into the Iterator (object → Rasterize (Sprite) → Iterator → many instances). For the wider picture on points-as-textures against continuously re-baked 3D looks, see Sprites, impostors & billboards.
The output type changes with the mode: Raster produces a field, Sprite a collection of points. An existing wire keeps working, because a points-to-field adapter splices in automatically. The downstream node sees an image either way, so watch for a conversion you did not intend rather than a broken wire.
| Parameter | Type | Default | What it does |
|---|---|---|---|
input | Color Field or Volume or Field or Primitives or Collection | — | Content to rasterize into pixels — shapes, geometry, a field, or a volume (slice it with a plane) |
mode | String | "Raster" | Raster: bake to a raster (data). Sprite: place the bake as a textured sprite in the 3D scene. One of: Raster, Sprite. |
face_camera | Boolean | true | On: the sprite faces the camera (billboard). Off: a fixed-orientation quad you can tilt. |
camera | Camera3D | — | Camera for capturing 3D content (optional; flat 2D content ignores it) |
plane | Transform | — | Slice plane for a field/volume — wire a Transform3D and drag to slice |
projection | String | "First" | Field slice: First (flat), Max (MIP), Min, Average, Sum over the slab. One of: First, Max, Min, Average, Sum. |
samples | Number | 1 | Slab depth samples (1 = flat slice; more = smoother projection) |
resolution_mode | String | "Canvas" | How the baked raster's pixel size is chosen. One of: Canvas, Custom, Scale, Fit Content. |
res_width | Number | 1024 | Baked width in pixels |
res_height | Number | 1024 | Baked height in pixels |
res_scale | Number | 1 | Multiple of the canvas size |
padding | Number | 0 | Transparent bleed added around the content (px) |
fit_padding | Number | 0 | 3D auto-frame margin as a fraction of the content size |
supersample | String | "1×" | Render at N× then downsample for clean edges. One of: 1×, 2×, 4×. |
format | String | "16-bit float" | Pixel format / bit depth. Float = HDR (needed for feedback / accumulation). One of: 8-bit, 16-bit float, 32-bit float. |
premultiply | Boolean | true | Store premultiplied alpha |
color_space | String | "Linear" | How baked pixel values are encoded. One of: Linear, sRGB. |
background | String | "Transparent" | What fills pixels not covered by content. One of: Transparent, Solid. |
background_color | Color | rgba(0.00, 0.00, 0.00, 1.00) | Solid background fill color |
reproject | String | "Live" | Live: re-bake every change. Baked: render once and freeze. One of: Live, Baked. |
bake_frame | Number | 0 | Frame stamp for the baked capture — change it to re-bake |
strength | Number | 1 | Effect strength / wet-dry (0 = no effect, 1 = full). Editable, wirable, keyframable. |
Rasterize and the Render node
Rasterize is a single-input content converter: point one piece of content at it and get an image back. The render node is the scene-wide renderer, with lights, materials, and extra output passes like depth and normals. Use Render when you need those passes. Rasterize deliberately keeps to one clean image rather than half-duplicating the renderer.