Hands-on: rebuilding China's Feihong Pagoda with GPT-6 Astra + Blender MCP from blurry photos
AI寒武纪 · wechat · 2026-09-06
A Chinese blogger tested GPT-6 Astra + Blender by rebuilding the Feihong Pagoda — the world's tallest glazed pagoda — using only blurry reference photos from short-video apps.
- The model auto-detected local Blender, Node and FFmpeg, then produced a 13-tier tower modeling script and a Three.js interactive viewer (3 viewpoints, layer expansion, lighting toggles) plus a first 20s 1080p loop video.
- On iteration two it self-diagnosed overly repetitive tiers and geometric Buddhas, rebuilt the component library (clothed figures with fingers and armor, dragons with scales and claws, five-layer dougong brackets, lotus niches and finial), compressed the web model to 19MB, and output a 32s video with close-ups.
- It then drove Blender directly via MCP to render full-tower and detail shots, gray-model and wireframe images, auto-packaging the PNGs.
- The author notes heavy quota burn at medium effort level (two reset cards), and cites Greg Brockman's launch-event claim that this may be OpenAI's AGI answer. Verdict: MCP-driven modeling feels limitless.
Note: claims unverified; original is a self-media hands-on post.
More from coding & agent
- Code Arena Launches WebDev Leaderboard Ranking AI Coding Models — arena · 2026-09-06
- Astra handles UV unwrapping, hinting at automating artists' grunt work — rms80 · 2026-09-06
- Macroscope ships Murmur, the in-house tool that let engineers direct dozens of cloud agents — Rasmic · 2026-09-06
- Astra disappoints on harness-building tasks while Fable 5.1 excels, dev reports — HarveenChadha · 2026-09-06
- Andrew Ng: prompting will be dead in 6 months, graphs are replacing it — nikola_mr64990 · 2026-09-06
- GPT-6 Astra designs a working jet plant and ships a live 3D simulation autonomously — deanwball · 2026-09-06