The viewport
The live window onto the scene: fly the camera, look through your own cameras, and switch on the overlays that help you compose.
Everything in DNA lives in one 3D-native space, flat 2D work included. The viewport shows the result of the graph and updates as nodes change.
Perspective and orthographic
Two projections, switchable at any time from the viewport toolbar. Orthographic is the default: flat and parallel with no perspective falloff, which suits laying out flat 2D work, lining things up, and precise placement. Perspective is the lens view, where things get smaller with distance, and it is what makes a 3D scene read as a real space.
Switching does not jump or re-frame the view. Orthographic still orbits, so it remains a full 3D view without the lens distortion. There is no separate 2D mode; the choice is a projection.
Moving the camera
Orbit with
Alt+drag, rotating around the point being looked at.Alt+double-click sets that point.Pan with a middle-mouse drag, with
Alt+Shift+drag, or with a two-finger trackpad drag.Zoom with the scroll wheel, around the cursor.
Alt+scroll drives the 3D camera instead of the canvas, andAlt+right-drag dollies.
Press F to focus the selection, so orbiting pivots around that object rather than the world origin. With nothing selected it fits the canvas instead.
Looking through a camera
The default is a free interactive camera used only for navigation. Any Camera node in the scene can instead be looked through, showing exactly what that camera sees, framing and lens included.
While looking through a camera node, orbiting, panning, dollying and zooming drive that node directly and are saved back onto it. Selection handles and picking stay locked to the rendered view. A camera in FPS mode enters free-look and captures the mouse when looked through; Esc releases it.
A Camera node is the viewpoint and lens only. Final pixels still come from the Render node. See Cameras.
Why a wired node stops drawing
Wire a cube into a node that consumes it, a Rasterize or a Transform, and the cube stops appearing on its own. Only the downstream result is drawn, which is what keeps a chain of five nodes from stacking five copies of the same thing on screen.
That is an authored value rather than a rule you cannot reach. Making the connection writes the upstream node's opacity to 0, and cutting the last wire out of it writes the opacity back. Drag that opacity back up to 1 and the cube returns, so the cube and the rasterised card are both on screen at once. Zero here is a hide rather than a fade: the node stays in the scene and whatever captures it still sees it at full strength. The Layer Manager lists every visual node whatever its opacity, so a node the wiring turned down is still there with its slider at the bottom.
An existing project acquires those zeros once, the first time it is opened, so nothing that already looked right changes. A node inside a subgraph behaves identically to one at the root.
Pixel Preview
Pixel Preview is a toggle in the viewport's Overlays menu, off by default.
Off, the composition is drawn at the screen's pixel density, so vector shapes, text and Analytic shapes stay crisp at any zoom. Use it for clean authoring.
On, the comp is rendered once at the project's export resolution and those pixels are magnified, so what is on screen is what exports. A 1:1 pixel grid appears once a comp pixel covers about 6 screen pixels. Use it to check aliasing and fine detail.
The dark grey border around the canvas is not a Pixel Preview cue. It is the off-canvas fill, drawn whenever the graph holds a Canvas node, in both modes, and it is not one of the overlays the master switch hides.
Overlays
Overlays are display-only and never change the scene, and none of them appear in exports. Show Overlays is the master switch that hides the lot.
Grid, a ground grid with coloured origin axes. It draws as lines; the grid style and subdivision settings in the data model are not read by the renderer.
Composition: Boundary, Safe Areas, Center Cross and Rulers.
Geometry: Wireframe and Normals, the mesh edges or hairs showing which way each surface faces, which is how flipped geometry gets found. Bounding Boxes, Point IDs, Object Names, X-Ray Mode and Random Colours for telling shards and pieces apart sit alongside them.
Debug: FPS Counter, Draw Time and Debug Info.
Three toggles in that menu draw nothing today. Motion Paths and Onion Skinning have painters but are never handed the samples they need, and Backface Tinting has no renderer behind it.
Snapping
The menu carries Always Snap, Snap to Grid, Snap to Guidelines, Snap to Objects, Snap to Vertices and Equal Spacing.
None of them affect placement today. The toggles write to one setting and the snapping code reads another, so a drag is never snapped whatever you switch on. There are no user guidelines to snap to either.