Open-source image-blaster turns a single photo into an explorable 3D world in 5 minutes
Open-source project image-blaster is trending on GitHub: given any single photo, it generates a fully explorable 3D world in about 5 minutes. Built as a Claude Code skill set chaining World Labs and FAL, it is MIT-licensed and, per reposts, has gathered 5.3k stars.
Confirmed
- Output includes 3D models of dynamic objects (.glb/.obj), physics-enabled meshes, static backgrounds reconstructed with 3D Gaussian Splatting, and automatically added ambient audio.
- The pipeline is orchestrated by Claude Code, almost fully automated, taking about 5 minutes end to end.
- In a hands-on test, the author fed in a 1984 East Berlin scene generated with GPT Image-2; per @IDRMcGregor's repost, the result was impressive. Per @jasonmayes, the original poster said it would shock even someone with ten years of Blender experience.
Why it matters
- The project demonstrates Claude Code's ability to orchestrate multiple AI services into an end-to-end workflow, dramatically lowering the barrier from a single image to a 3D scene, with direct value for games, film, and rapid prototyping.
2026-10-04 ~ 2026-10-05 · 5 related posts
Primary sources
- image-blaster Turns a Single Photo into an Explorable 3D World in 5 Minutes — jason_mayes · 2026-10-04
- image-blaster: open-source Claude skill turns one image into a 3D world in under 5 minutes — ID_R_McGregor · 2026-10-04
- Image-blaster turns one image into a 3D world in minutes, orchestrated by Claude Code — ID_R_McGregor · 2026-10-04
- [source] Open-source Claude skill turns one photo into an explorable 3D world in 5 minutes — Kanute3333 · 2026-10-04
1 near-duplicate retellings: Aiden_Tech_Ai