One evening, one 256MP image: isometric map of 12.66 km² of downtown Toronto drawn with a Qwen-Image LoRA
CSProfKGD · x · 2026-11-04
Developer John Fewell built "Isometric Toronto," an isometric illustrated map of downtown Toronto (Bathurst to the Don, Bloor to the harbour) rendered as a single 256-megapixel image covering 12.66 km² — about 2% of the city — drawn in one evening on a laptop, explorable with pan and zoom.
The pipeline streams Google Photorealistic 3D Tiles through an orthographic camera at 30°, registers them to Toronto open data, generates style illustrations with Gemini, trains a LoRA on Qwen-Image-2.1-Turbo to learn that style, then denoises the whole map as one image so there are no tile seams. The method follows Tiny London by Boris Starkov, building on Andy Coenen's Isometric NYC.
More from Multimodal
- AI-generated visual story of a brother's promise: sister's smile reflected in a horse's eye — LudovicCreator · 2026-11-04
- fal.ai lands back-to-back rounds: Sequoia-led Series E and Bessemer-led Series F with NVIDIA — reidhoffman · 2026-11-04
- Turning Suno tracks into lyric videos with Claude Code supervising Pokee Isaac — Kyrannio · 2026-11-04
- Creator Teases Timed-Lyric AI Video Following Instrumental Release — Kyrannio · 2026-11-04
- Electricity per AI image measured across RTX GPUs: ~3-10 cents per 1,000 images with FLUX.2 klein — strata2signal · 2026-11-04
- Topaz Labs launches Wonder 4, a one-click MoE-based upscale/sharpen/denoise model — aziz4ai · 2026-11-04