Canvas, safe-area and reel geometry specifications

Updated 2026-08-16

Delivered canvases

These are the normalized output sizes used by the pipeline. Provider render canvases may be a few pixels larger so their edges land on supported multiples, but integration should use the delivered files and the dimensions published in the manifest.

Do not infer placement from these dimensions. They describe each file, not where it belongs on a stage. Placement comes from the rectangles under the preview layout in asset-set.json. The production asset checklist covers the full handoff around those files.

Two stages are two compositions

The desktop background establishes the location. The phone background receives that image as an identity reference and redraws the same architecture, materials, weather, time of day and light for a vertical camera. It is not a 9:16 crop cut from the centre of the 16:9 art.

The distinction matters because the reel cabinet occupies most of a phone's width. A landscape landmark that lives outside the central band disappears in a crop, while a mascot gutter that works on desktop may not exist on a narrow stage. The portrait prompt reorganizes the location vertically: landmark or depth destination in the upper third, a calm middle band for the cabinet and a grounded foreground below it.

The reel-safe area is not a blank panel painted into the background. The environment continues through it with restrained contrast and no focal objects. The cabinet is composited over that environment as a separate opaque asset.

Five-by-three reel geometry

The nominal opening is computed from the stage rather than copied between orientations. It targets 56% of stage height, caps width at 86% of stage width and centres at 55% of stage height. Width and height are then held at exactly 5:3. Those ratios provide a fallback when no cabinet can be measured.

The generated cabinet contains a darker inner panel. After the cabinet is fitted, the pipeline walks from the centre of a grayscale probe toward the rails and detects the midpoint between panel and frame. A plausible result is fitted to the largest 5:3 rectangle inside the detected panel. If contrast, size or centring checks fail, the nominal opening wins. This is intentionally conservative: a nominal grid is better than symbols overlapping an incorrectly detected rail.

Each cell follows directly:

cell width = opening width / 5

cell height = opening height / 3

Because the opening is 5:3, those values match. An old 2.07:1 opening cropped square symbols in preview; preserving 5:3 removes that disagreement at the source.

Cabinet specification

The rail is one quarter of a reel cell on every side. Five cells plus two quarter-cell rails produce 5.5 cells across; three cells plus the same rails produce 3.5 cells down. The cabinet is therefore 11:7 on every stage. Its generation canvas is 1408×896, exactly 11:7.

Fit the cabinet uniformly into its published rectangle. Never assign independent display width and height: that converts round fittings into ellipses. One 11:7 source serves both stages because its proportions do not change.

Symbol fill and safe margins

The post-process measures alpha bounds and rescales the subject to its tier target. This corrects model renders that otherwise arrive anywhere from too small to edge-touching. It does not repair a confused silhouette, a cropped source or excess internal detail.

The 96-pixel readability test

1. Composite each symbol over the actual cabinet panel, not a white review page.
2. Display the complete cell at 96×96 without sharpening it for the test.
3. Identify the subject from its outer silhouette before inspecting internal marks.
4. Compare Wild and Scatter beside each other. Closed/external light and open/internal light must survive.
5. Compare one high with one low. The material depth and plaque should lift the high without making the low illegible.
6. Spin the reels. Fine details that appear static may merge in motion.

For the logic behind the tier budgets, see slot symbol tiers and paytable art.

Mascot and portrait framing

The mascot output is exactly 512×1536. Its alpha-cropped figure fills no more than 98% of the width and 93% of the height, with the baseline at 96.5% of canvas height. The figure includes a small piece of footing so the contact point survives any background. The 1:3 shape fits the gutter at close to full board height without entering the opening.

Portraits are different assets. They use 768×1024 canvases and may show a waist-up or full figure at 84–90% of canvas height. Their faces are judged at 256 pixels, while the mascot bust must remain identifiable at 96. Do not crop a portrait into a reel icon and call it the same specification. The slot mascot guide explains the identity handoff.

Manifest rules for integrators

• Read stage dimensions, background asset, board rectangle, reel window, title slot, mascot slot and corner flips from the relevant orientation.
• Use one uniform scale per asset inside its box.
• Keep base symbol and registered layers on the same origin.
• Do not patch a visual mismatch in the scene. If preview and client disagree, correct the published manifest geometry.
• On orientation change, reposition existing sprites against the second stage rather than treating portrait as a crop of desktop.

For bonus mode, preserve these same rectangles unless the game design has an explicit mechanical reason to change them. The bonus-mode art guide shows how to create contrast without breaking the base layout.