The render pipeline
Last updated
From composition to frame
Section titled “From composition to frame”A composition can be viewed as a document, with each layer explaining how to render something specific. Before rendering, Lumit compiles it into a DAG, or evaluation graph, which calculates everything required to render that particular frame.
For instance, identical layers or duplicates (such as duplicating a precomp) only need to be rendered once, and the output is reused. This helps reduce the work required when rendering expensive frames.
The DAG is completely internal. As a user you never interact with it or see it.
Content hashing
Section titled “Content hashing”Each item of work is hashed, which is altered by anything that affects the rendered result: the source layer, transform property changes, effects.
If two items hash as the same value, then they will be visually identical, which allows us to reuse the cached rendered texture.
Order of operation
Section titled “Order of operation”Within a composition, it renders from the bottom layer to the top. If a layer references another layer that is above it in the composition, or that wouldn’t be otherwise rendered (e.g. invisible), then it will stop rendering the current layer until the referenced layer is rendered, then continue.
For each layer, it renders in the following order: decodes source frame, retime, flow, masks, effects, transform, and then uses the blend mode to composite into the existing frame.
The effects and masks are composited from the top item to the bottom as they appear in the effects column or in the mask layer row. For instance if you apply a glow effect above (meaning before) an exposure effect, it will have a very different effect on the output, than if you reverse the order.