Skip to main content

Multiple Cameras

A scene can contain any number of cameras, and they compose into the final image according to a simple model: every enabled camera renders its layers into its viewport of its render target (the screen, unless a render target is set), in order of priority — lower values render first.

This model supports a wide range of setups: split-screen multiplayer, picture-in-picture overlays such as minimaps and rear-view mirrors, UI rendered over the 3D scene, and live render-to-texture surfaces like security monitors and portals.

Multi View

Viewports​

By default, a camera renders to the full width and height of its render target. The rect property restricts rendering to a rectangle, specified as [x, y, width, height] in normalized 0–1 coordinates (the origin is the bottom-left corner).

For 2-player horizontal split-screen, two cameras each take half the screen:

Horizontal splitscreen

// Player 1: left half of the screen
camera1.camera.rect = new pc.Vec4(0, 0, 0.5, 1);

// Player 2: right half of the screen, rendered after camera1
camera2.camera.rect = new pc.Vec4(0.5, 0, 0.5, 1);
camera2.camera.priority = 1;

For vertical split-screen, stack the viewports instead — [0, 0.5, 1, 0.5] on top and [0, 0, 1, 0.5] below:

Vertical splitscreen

A related property, scissorRect, clips rendering to a rectangle in the same normalized format without changing how the image is projected into the viewport.

Layers​

Each camera renders only the layers listed in its layers property, so different cameras can see entirely different subsets of the scene. Typical uses include a UI camera that renders only a UI layer over the game, a minimap camera that skips effects layers, and first-person weapon rendering. See the Camera Model Masking tutorial for a worked example.

Camera Stacking​

When one camera renders on top of another — a picture-in-picture overlay, or a full-screen camera drawing a different set of layers — the later camera (higher priority) must not wipe out the earlier camera's image. Disable its clear flags as appropriate:

// Render an overlay camera on top of the main view, in the bottom-right corner
overlay.camera.priority = 1;
overlay.camera.rect = new pc.Vec4(0.7, 0, 0.3, 0.3);
overlay.camera.clearColorBuffer = true;  // overlay has its own background
overlay.camera.clearDepthBuffer = true;  // don't depth-test against the main view

// For a full-screen overlay that composites over the main view instead:
// overlay.camera.clearColorBuffer = false;

Render Targets​

Instead of the screen, a camera can render into an offscreen texture by assigning a RenderTarget to its renderTarget property. The resulting texture can then be applied to a material — for in-world screens, mirrors and portals — or processed further. See the Render Targets page for the full details, and the engine's render-to-texture example:

Render to Texture

Performance Considerations​

  • Every enabled camera renders its layers again — draw calls scale with the number of cameras. Restrict each camera's layers to the minimum it actually needs.
  • Reducing a camera's rect reduces the pixels it fills, but not the per-draw-call CPU cost of the objects it renders.
  • Per-camera post-processing runs once per camera, so effects on split-screen views multiply their GPU cost.
  • Render targets that are only needed occasionally (e.g. a static mirror) don't have to be updated every frame — disable the camera and enable it on demand.