The Social Media Template Kit is 100% client-side. Each layer — a text block, a rectangle, an ellipse, a line or an uploaded image — is a small typed object of numbers and strings held in the page, and a pure engine (socialKitEngine.ts) turns those objects into pixels on an HTML5 <canvas> on your device. When you upload a photo it is read straight into a browser object URL with URL.createObjectURL, decoded into an in-memory image and composited on the canvas locally; it is never uploaded or transmitted anywhere. The same drawSocial() routine paints the live preview and the export, so what you see is what you download: the PNG is produced with a local canvas.toBlob at the exact platform pixels, and the “all sizes” download is a ZIP assembled in memory with JSZip. Fonts are your device’s own system fonts, so nothing is fetched, and there is deliberately no built-in stock or AI imagery because that would require a third-party server. The tool works offline once cached; close the tab and anything you did not export is gone.

This kit ships each platform’s recommended dimensions as of July 2026. Most feeds are happiest at 1080 px wide: Instagram uses 1080×1080 for square posts, 1080×1350 for 4:5 portrait (which now tends to outperform square) and 1080×1920 for a 9:16 story or reel. Facebook link posts are about 1200×630, its story is 1080×1920 and the page cover is 820×312. X/Twitter in-stream images look best at 1600×900 (16:9) and the profile header is 1500×500 (3:1). LinkedIn link posts are 1200×627 and the page cover is 1584×396. YouTube thumbnails are 1280×720 and channel art is 2560×1440 with a central 1546×423 safe area shown on every device. Pinterest pins are 1000×1500 (2:3) and TikTok is fully vertical at 1080×1920. Platforms change specs occasionally, so verify the current spec before a critical upload — and keep key text near the middle because stories, reels and channel art crop differently on each device.

The design is stored on a fixed-width virtual artboard, so switching platform size only changes the canvas height. Magic-resize preserves each layer’s horizontal layout and text size and re-positions layers so they keep the same relative vertical placement on the new canvas, instead of clipping. That turns one design into a whole size set in a click and is a genuine time-saver — the same engine even batch-renders every ticked size into a ZIP. It is a starting point, not magic: extreme aspect changes (a tall 1080×1920 story versus a wide 1500×500 header) leave big gaps or crowding, so you usually nudge a few layers per size. Because the preview and export share one render routine, any tweak you make is exactly what downloads.

Text renders in system font stacks so the tool works fully offline and never fetches a webfont; the trade-off is that a font can render a little differently on a machine that lacks it, and there is no embedded typeface in the exported PNG. There is deliberately no stock-image library and no AI image generation, because both would require a third-party server and would break the promise that files never leave your device. Instead you upload your own photo (read locally and composited on-device) and build backgrounds from solid colours, linear or radial gradients, or procedural inline patterns (dots, grid, diagonal, triangles, checker) that stay crisp at any export size. Exports are flat RGB PNGs held in memory, so export before you close the tab.