Canvas, Channels & Layers

PaintCanvas is an abstract lifecycle host — it resolves the Paintable target, owns the UV-seam environment, and drives the whole system by broadcasting lifecycle phases (Initialize, Update, Reset, Clear, SourceChanged) down through its channel/layer tree every frame. It also implements IDrawContext directly, so a PaintDrawer resolves its stamp targets straight from the canvas — there is no separate per-canvas context object. Three concrete canvases build on it.

The three canvas types

Canvas	Best for	Notes
MultiChannelCanvas	The general case — several channels, each with its own layer stack	Populates the shared ChannelTopology that groups channels by shader property
SimulationCanvas	Physically simulated wet paint	Drives a shared SimulationWorkspace; needs a sibling fluid solver — see Fluid Simulation
SingleTargetCanvas	One channel, no layer stack, cheapest option	A stamp lands straight in its own buffer; one publish draw per dirty frame composites it over the material — no layer stack, no scratch, no shared topology

Channels (ChannelDefinition)

Each channel is authored as a ChannelDefinition asset:

Field	Description
Value Type	Color, Scalar, or Normal
Shader Property / Keyword	The target material property the channel drives
Default Value	Fallback value used before anything is painted
sRGB / Bit Depth	Colour-space and precision settings for the channel's buffers

Because the shader property is fully configurable, a channel can drive albedo, a metallic/smoothness mask, a normal map, or any other property your material exposes.

Layers (PaintLayer)

On a MultiChannelCanvas or SimulationCanvas, each channel owns an independent stack of layers (PaintLayer): visibility, opacity, an optional starting texture, and a blend mode. Blend modes are tracked separately per value type, so switching a channel's type never misinterprets a setting that belonged to a different family:

Value type	Blend modes
Color	Normal, Multiply, Add, Min, Max, Screen, Overlay, Soft Light
Scalar	Normal, Multiply, Add, Min, Max
Normal	Lerp, RNM (Reoriented Normal Mapping), UDN, Whiteout, Overlay, Max Slope, Subtract

A SingleTargetCanvas skips layers and compositing entirely — a stamp lands straight on its one buffer, which is why it's the cheapest of the three.

Compositing

Every frame a dirty channel is composited: the origin texture, then all visible layers (bottom to top, each with its blend mode and opacity), then the in-progress scratch stroke, producing the final texture set on the material.

Retargeting at runtime

A canvas paints exactly one Paintable at a time. Call SetTarget(...) to point it at a different object — the existing render textures are reused rather than reallocated:

// Retarget the canvas onto a different paintable object.
paintCanvas.SetTarget(otherPaintable);

// Restore or wipe the current canvas.
paintCanvas.ResetToOrigin(); // back to each layer's starting texture
paintCanvas.Clear();         // wipe to the default background

---

Next: Input & Stroke Methods