image-blaster: Turn Images into 3D Environments and Assets

Overview

image-blaster is a Claude-driven workflow for turning a reference image into a collection of 3D and audio assets. It combines World Labs for an explorable environment, FAL for object meshes, and sound generation for ambient and object-specific effects.

The project is aimed at speeding up early 3D work, not replacing a full modeling or production pipeline. Its outputs can be used as starting assets in game engines, DCC tools, or web applications.

Key Features

• Generates a Gaussian splat environment from a source image.
• Creates meshes for dynamic objects in .glb or .obj formats.
• Produces ambient loops and object-specific sound effects.
• Uses Claude skills to coordinate the image-to-assets workflow.
• Supports Hunyuan 3D options for face count, PBR materials, geometry style, and polygon type.
• Supports image-editing preferences including nano-banana and gpt-image-2.
• Designed for integration into engines and tools such as Unity, Unreal, Godot, Blender, and Three.js.

Use Cases

• Game developers can turn a concept image into a rough level environment and object assets to begin prototyping.
• 3D artists can use a photo or illustration as a jumpstart for environment blocking and asset creation.
• Film and architecture teams can create quick spatial concepts from location or rendering references.
• Web developers can generate scene assets to explore interactive 3D experiences in a Three.js or Electron project.

Project Facts

• Language: TypeScript
• License: MIT
• Stars: 9.4k
• Forks: 918
• Archived: No

Getting Started

Clone the repository, install Claude, configure World Labs and FAL API keys, then place an image in input/ and ask Claude to process it. See the README for the full setup and workflow.

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

Alternatives

• GigaSLAM: GigaSLAM reconstructs and tracks large outdoor scenes from RGB video, while image-blaster turns an image into 3D assets and sound effects.

Considerations

• Using the workflow requires Claude and API access to World Labs and FAL; generated assets depend on these external services.
• The outputs are intended to jumpstart 3D work, so they may need review and further editing before production use.
• The README describes model choices and generation parameters, but does not specify a local-only or offline workflow.