Perspective Lab
Perspective Lab at knowperspective.art is a free study tool for artists learning to construct linear perspective by hand. It renders the classic folded-plan construction — horizon, vertical through the centre of vision, station point, cone of vision, 90° legs, and vanishing points — and lets you push any of them around. Every line comes from a single pinhole projection of a real box through a real eye point, so the construction overlay and the drawing can never disagree.
Problem
Perspective books show the construction as a finished diagram. Artists can copy it but still can't say why each line is there or how the elements depend on each other — why moving the station point spreads the vanishing points, why turning the box pulls one VP in while the other runs off the sheet, or where a third VP comes from.
Approach
Built one projection model and derived every element on the sheet from it rather than drawing each line independently. Sliders drive the viewing distance, cone angle, box rotation, eye height, and tilt; the canvas re-renders the construction and the box together, with a hover-to-highlight element list that explains what each line is for and what moves it. One-, two- and three-point fall out of the same system: turn the box square for one-point, tilt the line of sight for three-point.
Key Features
- Live 1, 2 and 3-point construction from a single pinhole projection
- Hover any sheet element to light it up and read what it does and what moves it
- Plan-method overlay: the box plan folded onto the sheet with visual rays from the station point
- Camera-lens mode mapping focal length and sensor size to viewing distance and cone of vision
- 'Find D from the cone' helper that mirrors the paper method
- Presets, live readings and warnings, pan/zoom, and a quick-reference formula table
Outcomes
- Construction overlay and drawing are always consistent because both come from the same projection
- Static site with no build step and no tracking — loads instantly and runs entirely in the browser
- Ships with llms.txt, llms-full.txt, FAQ and WebApplication structured data for AI and search discovery
Architecture
Static HTML, CSS, and vanilla JavaScript with no build step. A single ~500-line module holds the projection model and the canvas renderer. Deployed on Vercel with clean URLs, a www-to-apex redirect, llms.txt, sitemap, and schema.org JSON-LD. Open source on GitHub.