New on my GitHub: Ruang.

Room photo to priced 3D plan: 7 render styles, drag-to-arrange, country renovation templates, local-currency estimate and an architect that fixes what it flags. Runs in the browser.

In short:

  • Will it fit through the door? – checks a new item's smallest cross-section against your door.
  • Walk mode – explore at eye height (1.6 m) with WASD or the on-screen pad.
  • Region + language first. 24 regions, 7 UI languages (English, Bahasa Indonesia, 繁體中文, 简体中文, 日本語, 한국어, Español). Currency, units, climate tips and the sun clock follow the region.
  • Walking-space heatmap – grid flood-fill from the door shows reachable floor and the narrowest bottleneck in centimetres.

Open source, so feel free to break it and tell me how.

What else it does

  • Look. Seven render styles: realistic, miniature (tilt-shift), pixel, cartoon (toon + ink outline), clay, blueprint and sketch. Photo-3D mode projects your own photo onto the walls.
  • Arrange. Drag, rotate, resize, duplicate, delete, undo/redo, 22 catalogue items. Live warnings for overlaps, blocked doors and windows, narrow walkways and door swing.
  • Sun clock + direct-sun hours – NOAA solar position for your city and date, shadows cast through your real window openings, and hours of direct sun per piece of furniture (glare on the TV, fading on the sofa).
  • Auto-arrange – simulated annealing over a layout score (clearances, light, door swing, focal walls).
  • Best look for my money – a knapsack budget optimiser picks the upgrades with the biggest visual impact that fit your budget.
  • Rent-safe mode – renovation plans that use only removable changes (rugs, lamps, peel-and-stick, plants, furniture) for tenants.
  • Culture lens – optional feng shui or vastu notes, shown as tradition, not science.
  • Room depth is not visible in a single back-wall photo; Ruang assumes the camera stood 0.4 m inside the front wall and lets you edit it. On the bundled sample the default depth is 4.55 m against a true 4.2 m.
  • Reconstruction assumes a rectangular room and a level phone; strongly tilted photos give worse sizes.
  • AI auto-detect downloads a model from Hugging Face on first use; it could not be tested on the build network, so treat it as experimental.

Under the hood

  • Written mostly in JavaScript
  • Hosted for free on GitHub Pages
  • Tests run in GitHub Actions on every push
  • MIT licensed

Links