Paste-Up
Compose the view as a collage instead of faking a photoreal render. Import images and cut them out with polygon scissors (hard edges by design, or a torn edge), lay flat colour fields from a curated post-digital palette, set type in Archivo Black, and stack it all as layers with blend modes and honest print-shop filters. Runs entirely in your browser — images never leave your machine.
Or drag & drop image files straight onto the board. Select an image layer to cut it with the polygon scissors (in the inspector).
Eye height is the horizon: a figure exactly as tall as the eye height has its head on the horizon line, at any depth. Switch the ground on and the board enforces that — drag a figure back and it shrinks by the rule, not by eye.
A plausible gradient keyed to the sun angle — not a physical sky model. Scattering-based skies (Hosek–Wilkie) are a separate tool, Firmament. It replaces the flat background colour while it is on.
A project file is JSON with your images embedded as data URLs — fine for a working file, but it can get large (a warning appears past ~8 MB). Write one from Export in the header. Nothing is uploaded either way.
Drop an image, or add a shape to start.
Nothing pasted up yet. Import an image or add a shape — layers stack here, top of the pile first.
Select a layer to edit it — drag to move, corner handles to scale, the stalk to rotate (Shift snaps 15°). Arrow keys nudge; Delete removes; ⌘/Ctrl-Z undoes.
Field guide — the post-digital moves
The look this tool serves — the collage renders of practices like Fala Atelier, OMMX and Point Supreme — refuses photorealism on purpose: the drawing admits it is a drawing, so every element reads as a decision rather than a camera accident. Its recurring moves have names.
The flat cutout. Imagery is scissored along a visible polygon — the hard edge is the point, not a masking failure. Cut people, furniture, trees and whole façades and paste them flat.
Oversized entourage. A plant, a curtain, a dog or a figure is scaled past realism so occupation — not construction — becomes the subject of the view.
The flat colour ground. Sky and floor become fields of one colour — dusty pink, sage, cobalt — so depth is composed, never simulated. The background picker and shape layers are for exactly this.
Visible texture. Halftone dots, grain and posterized tone declare the image is printed matter, not a window. Apply them per layer so a photograph sits comfortably next to flat colour.
Siblings on this bench: screen a whole image into riso plates at Halftone & Riso, and pull scalable silhouettes to cut and paste from the Entourage Library.
How it works
The board is a plain HTML canvas at a fixed working resolution (~2000 px on the long edge). Each layer renders to its own offscreen canvas at the source's native size, then is stamped onto the board with a centre/scale/rotation/flip transform; blend modes are the canvas compositor's own multiply / screen / overlay operations, so what you see is exactly what exports.
The scissors are a polygon clip path — your clicked vertices, mapped into the image's own pixel space, become a ctx.clip() region, which is why the edges stay hard at any scale. The torn edge is the same polygon with each edge subdivided every ~14 px and jittered along its normal by a seeded random amount, so a tear is reproducible across renders and exports. There is no auto-masking, magic wand or AI in here — v1 cuts only where you click, which is the pedagogical point.
Filters are straightforward pixel operations on the layer's offscreen canvas: posterize quantizes each RGB channel to 2–8 levels; desaturate and tint work on Rec. 601 luminance (tint maps shadows to the ink colour and highlights to white, like a one-ink print); grain adds seeded ±noise. The halftone is a real AM screen: the layer is sampled in rotated cells and replaced by dots whose area tracks the average darkness under each cell — the same model as the Halftone & Riso tool, applied per layer. Undo/redo keeps snapshots of the layer metadata (images are held once, by reference). Everything — imports, edits, PNG export, project save — happens in your browser; no image is ever uploaded.