Skip to content
Get packs

image-blaster

Turns one image into 3D models, a 3D environment, and sound effects using Claude skills.

GitHub Download source
Stars
12.3k
Forks
1.3k
Licence
MIT
Language
TypeScript
Last update
6 months ago
Type
AI model / agent

What it is

image-blaster is a set of skills for Claude, Anthropic's AI assistant, that takes a single input image and generates 3D content from it. According to the README, it uses World Labs and FAL services to create 3D object models, a Gaussian splat of the static environment, and looping ambient and object sound effects. The project is aimed at people who want a starting point for 3D environments quickly.

  • Generate a 3D environment from a photo of a room or location
  • Create .glb or .obj models of dynamic objects found in the image
  • Produce looping ambient sound and object-specific sound effects
  • Embed the output in a game engine, DCC tool, or web app, as the README suggests

Great for

  • Game developers who want level concept environments from a reference image
  • 3D artists who need a quick base to build on
  • Filmmakers or location scouts exploring visual ideas
  • Developers who already have Claude and World Labs or FAL API keys

Not ideal for

  • People without API keys for World Labs and FAL, since the README requires them
  • Anyone needing production-ready assets without review, because the README presents output as a jumpstart rather than a finished result
  • People unfamiliar with command-line tools and git, since setup starts in a terminal
Step-by-step guide: how to use image-blasterInstall, first run, examples, tips and FAQs

How to get it

  1. 1

    Clone and build from source

    You'll need: Git · Claude Code CLI (installed via the curl command in the README) · World Labs API key · FAL API key

    Clone the repository

    git clone https://github.com/neilsonnn/image-blaster

    Enter the directory

    cd image-blaster

    Install Claude (per README)

    curl -fsSL https://claude.ai/install.sh | bash

    Start Claude in the project

    claude
  2. 2

    Run it

    • Start Claude from inside the image-blaster directory with the claude command.
    • Give Claude your World Labs and FAL API keys when prompted.
    • Place an image into the input/ directory.
    • Ask Claude to blast it and confirm each step with you.
  3. 3

    Get the most out of it

    • Use the --face-count option to set target face count between 40000 and 1500000; the default is 50000.
    • Use --generate-type LowPoly or Geometry to change the output model style, per the README.
    • Use --polygon-type quadrilateral with LowPoly if you prefer quad meshes.
    • Embed the output in Unity, Unreal, Godot, Blender, or a Three.js app, as the README describes.
    • Remove /app from .claudeignore if you want Claude to modify the React viewer, per the Development section.

Downloads

Source code (.zip)main branch · straight from GitHub

Strengths

  • Bundles several generation steps (environment, meshes, audio) behind one image input
  • Uses the Claude skill format, so you drive it through a conversation
  • Output formats such as .glb, .obj, .spz, and .mp3 are named in the README

Trade-offs

  • Depends on paid third-party services (World Labs, FAL) whose costs and terms are not covered in the README
  • The README's under-five-minute claim is not independently verified and results will vary by image

Stats synced from GitHub. The assessment above was drafted with AI from the project's README on 2026-10-29 and may miss recent changes — check the README before relying on a command. MIT licence: free to use, modify and distribute, including commercially, as long as you keep the copyright notice.