Skip to content
Get packs

Guide · Open source

How to use image-blaster: set up, run and get 3D output

Updated · written with AI from the project's own documentation and checked against our catalogue data

image-blaster is an open-source set of skills for Claude, Anthropic's AI assistant, that takes one input image and generates 3D content from it. According to its 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. This guide walks through the setup and the basic run so you can get from an image to first output.

image-blasterneilsonnn/image-blaster · ★ 12.3k · MIT

Quick start

  1. 1Clone the repo. Run git clone https://github.com/neilsonnn/image-blaster in a terminal.
  2. 2Enter the folder. Run cd image-blaster to move into the project directory.
  3. 3Install and start Claude. Install Claude with the curl command from the README, then run claude in the project.
  4. 4Add API keys. Give Claude your World Labs and FAL API keys when it asks.
  5. 5Add an input image. Place your image in the input/ directory.
  6. 6Run the blast. Ask Claude to blast it and confirm each step with you.
  7. 7Review the files. Check the .glb, .obj, .spz and .mp3 files and refine them in your own tools.

What you need before you start

image-blaster is set up from a terminal, so you need Git and the Claude Code CLI. You also need your own API keys: the README requires a World Labs API key and a FAL API key.

Without those keys the project will not run, so get them before you clone the repository.

  • Git
  • Claude Code CLI (installed with the curl command in the README)
  • World Labs API key (platform.worldlabs.ai)
  • FAL API key (fal.ai)

Install image-blaster from the repository

Installation is clone-and-build. Open a terminal, clone the repository, and enter the project folder.

Then install Claude with the command from the README and start it inside the project.

git clone https://github.com/neilsonnn/image-blaster cd image-blaster curl -fsSL https://claude.ai/install.sh | bash claude

Give Claude your API keys

When Claude is running, say hello and give it your World Labs and FAL API keys. The README asks for both at this step, so the skills can call those services.

Keep your keys private and do not commit them to the repository.

Run a blast on your image

Put an image into the input/ directory. Then ask Claude to blast it and confirm each step with you, using the wording from the README.

The README describes the whole process as taking under 5 minutes to get from an image to a fully meshed 3D environment. Treat that as the project's claim, not a guarantee for your image.

blast it and confirm each step with me

Know what files you get back

By default image-blaster creates three kinds of output from your image. Dynamic objects become 3D models in .glb or .obj format. The static environment becomes a Gaussian splat in .spz format. Ambient looping sound and object-specific physics sound effects are saved as .mp3.

  • 3D models (.glb, .obj) of dynamic objects
  • Gaussian splat (.spz) of the static environment
  • Ambient looping sound and object SFX (.mp3)

Adjust the 3D model settings (optional)

The README lists Hunyuan parameters for 3D model creation. Use them only if you want to change the defaults. The face count default is 50000, while Hunyuan's own API default is 500000.

The generate type controls whether the model is textured, polygon-reduced, or a white geometry-only mesh.

  • --face-count <40000-1500000>: target face count, default 50000
  • --enable-pbr true|false: PBR material generation, default true
  • --generate-type Normal|LowPoly|Geometry: default Normal
  • --polygon-type triangle|quadrilateral: for LowPoly, default triangle

Review the output before using it

The README presents the output as a jumpstart rather than a finished result. Plan to review and rebuild assets in your own tools before shipping anything.

You can also embed the output in Unity, Unreal, Godot, Blender, Maya, 3DS Max, Three.js or Electron, according to the README's Extensions section.

What people use it for

  • Game developers building level concept environments from a reference image
  • 3D artists who need a quick base model to build on
  • Filmmakers or location scouts exploring visual ideas
  • Architectural rendering exploration, as the README's examples suggest
  • Creating a starting environment for a robot scene

Frequently asked questions

Is image-blaster free to use?

image-blaster is released under the MIT licence, as listed on its repository. The World Labs and FAL services it calls require your own API keys, so check their terms and any costs on those services.

What do I need to install image-blaster?

You need Git, the Claude Code CLI, a World Labs API key and a FAL API key. Setup starts from a terminal with git clone and cd image-blaster.

Can I use image-blaster without a World Labs or FAL API key?

The README requires both keys, so the project is not usable without them. You need to obtain them from platform.worldlabs.ai and fal.ai.

What file types does image-blaster output?

By default it creates .glb and .obj 3D models for dynamic objects, a .spz Gaussian splat for the static environment, and .mp3 sound files for ambient and object effects.

Is the output ready for production games or films?

No. The README presents the output as a jumpstart for 3D work, so it should be reviewed and refined before production use.

Are there alternatives to image-blaster?

The project's analysis lists no comparable projects, so this guide cannot name direct alternatives. You can still use general 3D tools such as Blender to refine the output.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

image-blaster suits developers and artists who already have Claude, World Labs and FAL API keys and want a fast starting point for 3D environments from one image. It is a poor fit for anyone who needs production-ready assets without review, or who is uncomfortable working in a terminal and with git.

Install options & downloads