The render node

Gathers everything in front of the camera, lights it, and hands back an image.

The Render node is the single crossing point from a 3D scene to flat pixels. Cameras are viewpoints and lights are lights; nothing becomes an actual image until a Render node makes one. Feed it a scene, choose a camera, and it returns a finished frame to composite, grade or export.

What it does

Each cook walks the whole scene once. Every solid object is drawn, shadows are cast, ambient occlusion and any reflections are added, and everything is lit in one pass. The see-through things follow, layered over the lit result: volumes, Gaussian splats, glass and refraction, glowing and unlit surfaces. Post effects run after that, and the bright, wide-range image is tone-mapped down to something a display can show. Flat 2D content and the viewport overlays composite after the tone map, so they keep the colours they were authored in.

That order is worth knowing, because it explains why a transparent object always lands on top of the solid scene behind it, and why Bloom and Depth of Field apply to the finished frame.

Leaving include unconnected captures the whole scene, which is what most graphs want. Wire geometry into it and that geometry is consumed: it leaves the live scene and appears only on this render's plane.

Choosing a camera

Leave the camera input empty and the active viewport camera is used, so the render follows however you orbit and zoom. Wire a Camera in and the render locks to that viewpoint instead, which is what a fixed shot, a second angle or an export needs.

See Cameras for setting up and animating cameras, and Choosing a renderer for picking which renderer draws the scene.

Output passes

Beyond the final lit colour, the node hands out eight more image passes, all computed along the way:

• depth, normals and position, the building blocks for fog by depth, relighting and masking by position, without re-rendering. normal_space puts normals in world, view or object space; position_range and the three position_offset_* dials map world units into the colour range.

• albedo, ao and emissive, straight off the G-buffer.

• object_id, a raw integer id per object rather than a colour image. Key keys on it directly.

• velocity, raw screen-space motion vectors rather than a colour image.

Eight of the nine are extracted every cook whether or not anything reads them, so wiring one costs nothing extra. velocity is the exception: its pass runs only when the pin is connected or motion blur is on.

There is a tenth pin, settings, and it is not an image. It carries this render's post-process policy out of the cook, which is how the viewport knows which effects to run over the frame it is showing.

Motion Field turns velocity into the same flow map Optical Flow produces, so a rendered scene and a piece of footage drive Displace and the other motion tools through one representation. A 3D scene knows its own motion, so nothing is estimated and there is no reference frame to set. With velocity unconnected the node reports no motion rather than a stale frame.

pass_type picks which of colour, normals, depth or position also comes out of the primary color pin, for when you want to look at a pass in the viewport rather than wire it away.

Ray-traced lighting

The ray-tracing stack is not on this node. It lives on Render Settings, which is one per composition and has no output pin: drop it in the graph and every render finds it, the same way the Canvas and Timeline nodes work.

rt_enabled is on by default. Rather than approximating shadows and contact darkening with screen-only methods, rays are traced into the actual scene, so shadows land accurately, creases darken where surfaces really meet, and reflections pick up things a screen-only method would miss. It stays interactive by reading what is already on screen first and tracing rays only for what that fast pass cannot see.

rt_shadows and rt_ao are on from the start. Two heavier tiers are off because they cost more: rt_reflections makes glossy and metallic surfaces mirror the scene, reading the visible frame first and tracing rays to fill in whatever is off-screen or hidden, and rt_gi is indirect bounced light spilling colour from one surface onto its neighbours.

Two dials keep the cost in hand:

• rt_quality traces rays at Full, Half or Quarter of the output resolution. Half is the default: it traces a quarter as many rays and rebuilds crisp edges from the full-resolution depth and normals, usually indistinguishable from full at a fraction of the cost.

• rt_shadow_budget caps how many lights cast traced shadows in a single frame, at 4. The strongest lights, and any directional light, always keep their shadows; the rest drop out in order of brightness. Set it to 0 to let every light cast.

Render Settings also carries the integrator choice between real-time hybrid tracing and the path tracer. See Choosing a renderer for that, and Making it faster for keeping a heavy scene interactive.

Parameters

See also

• Choosing a renderer

• Cameras

• Lights

• Post effects