Everyday 2D nodes

Regular 2D nodes are the default language of Fusion compositing. They receive an image, modify it and pass another image downstream. MediaIn, Background, Merge, Transform, Blur, Glow, Color Corrector, keyers, matte tools and many generators all live in this world.

The important distinction is the data on the wire: it is already raster image data. A 2D node can work with pixels, RGB, alpha, auxiliary channels, resolution and the image domain, but it does not understand a particle simulation or a 3D scene as a live scene.

Shape nodes pass vector-shape data until SRender converts it to an image. Particle nodes pass a particle stream until pRender (2D) turns that simulation into pixels. Classic 3D nodes pass a scene until Renderer3D renders it. USD/U nodes use their own scene-description pipeline and renderer/output stage.

Once those systems have been rendered, their results can be treated like any other image: merged over footage, blurred, graded, masked, given glow, sharpened and finally sent to MediaOut. This is why the regular 2D pipeline becomes the final meeting point for many different Fusion systems.

The color and role of an input matter. A main image input expects image data. Merge has separate Background and Foreground image inputs. A blue effect-mask input usually limits where a node operates; it does not replace the image that node processes.

That distinction explains many connection errors. If you want to pass a visible layer, use an image input. If you only want to constrain the effect, use the mask input. If the upstream branch still contains Shape, Particle or 3D scene data, render that branch before feeding it into the normal 2D composite.

Keep logical branches separate: source preparation, graphics, masks, effects and final assembly should be readable as blocks. A good Flow makes it obvious where an element is created, transformed, masked, composited and finally delivered.

Also watch the image domain. Heavy Blur, Glow and Transform operations can become unnecessarily expensive when the active domain is much larger than the visible element. Crop, SetDomain and Resize can be production tools, not merely housekeeping nodes.

A common graphics chain is MediaIn for the shot, Text+ or Background for the graphic, Transform for size and position, Merge for compositing, an optional mask for local control, then Blur, Glow or Color Corrector for finishing before MediaOut.

If particles are needed, build them separately as pEmitter → pTurbulence → pRender (2D), then merge the rendered image. If a 3D title is needed, build Text3D + Camera3D + SpotLight → Merge3D → Renderer3D, then return that rendered image to the 2D Merge.

Use regular 2D nodes whenever the job is already expressed as an image: placing a title over footage, correcting color, adding glow or blur, combining layers with Merge, preparing alpha, building a clean plate or sending the finished image to MediaOut.

This is also the most practical space for integration work. Screen replacements, logos, lower thirds, before/after composites, grain matching and final color balancing are usually easiest once every source has been converted to ordinary image data.

Even when a comp begins with Shape, particles or 3D, the finishing stage normally returns here: render the specialized system, merge the resulting image, match color and softness, verify alpha and deliver the final 2D composite.

Merge receives two independent images: MediaIn → Background and artwork → Foreground. Rectangle or Polygon arrives through a separate branch at Effect Mask. That connection controls the blend; it is not inserted in series between footage and an effect. View a mask directly to inspect its coverage signal, rather than expecting a colored title.

The Background input establishes the basic output format of Merge. Swapping full-frame footage with a small still may change dimensions, cropping and placement. Check Width/Height, Pixel Aspect and the current frame on both sources before adding more Transform corrections.

Inspect RGB and Alpha separately, then view the layer over both light and dark backgrounds. A black Viewer does not distinguish transparent pixels from an opaque black image. If alpha is one everywhere, the foreground will cover the background regardless of its RGB values.

For a nonlinear grade on premultiplied imagery, use an intentional AlphaDivide → grade → AlphaMultiply pair or the relevant tool’s built-in Pre-Divide/Post-Multiply. Do not apply both corrections automatically: multiplying twice darkens semitransparent edges. Recheck Z and ID channels after operations that rebuild the image.