Aspect ratios

One scene can render every platform ratio in a single pass. You author once, and the same source resolves to a landscape hero, a vertical Reel, and a square post.

./bin/vawe formats/scene/my-video.json --aspect 16:9,9:16,1:1
# → my-video.16x9.mp4, my-video.9x16.mp4, my-video.1x1.mp4

Supported ratios: 16:9, 9:16, 1:1, 4:5, 4:3. Each output is tagged with its aspect. Through make:

make video D=formats/scene/my-video.json ASPECT=16:9,9:16,1:1

Why it works: relative coordinates

For a scene to survive being reshaped, its layers cannot be pinned to fixed pixels. Three tools let a layer describe where it belongs, and the engine resolves that to pixels once per aspect, at boot, purely from the width and height.

Relative values. x, y, w, and h accept a pixel number, a percentage ("50%"), a percentage with an offset ("50%-40"), or a keyword (center, left, right, top, bottom) anchored inside a per-aspect safe inset.

{ "type": "text", "text": "Centered anywhere.", "x": "50%", "y": "40%" }

Pins. pin places a layer canvas-relative within the safe zone, including rule-of-thirds power points (thirds-tl, thirds-tr, thirds-t, and so on). Give the layer a w and h so the centering math knows its size.

{ "type": "text", "text": "Bottom third.", "pin": "thirds-b", "w": 900, "h": 200 }

The 12-column grid. col places a layer on a grid: "3" is a single column, "2-7" spans columns 2 through 7 and sets both x and width. cols (default 12) and gutter tune it.

{ "type": "image", "src": "photo.jpg", "col": "7-12" }

Determinism holds

Resolution happens once, at boot, as a pure function of the canvas dimensions. So a multi-aspect render is still fully deterministic: each ratio is its own reproducible output. Scenes authored with absolute pixel coordinates are unaffected; they simply render at their fixed positions.