Summary

image-blaster is an open-source tool designed to rapidly generate 3D environments, meshes, and sound effects from a single input image. Targeting game developers, 3D artists, and researchers, it leverages Claude AI skills, World Labs, and FAL to automate asset creation, aiming to transform an image into a fully meshed 3D environment in under five minutes, significantly accelerating 3D project workflows.

How It Works

The project orchestrates a pipeline using Claude AI skills to interpret an input image and delegate tasks to specialized generation models. World Labs' marble-1.1 model constructs the explorable environment, while FAL's hunyuan-3d model generates detailed 3D object meshes. Complementing the visuals, elevenlabs-sfx creates ambient and object-specific sound effects. Image cleanup and reference generation are handled by nano-banana or gpt-image-2. This integrated approach automates complex 3D asset generation, offering a novel way to quickly populate scenes from minimal input.

Quick Start & Requirements

To begin, clone the repository (git clone https://github.com/neilsonnn/image-blaster) and navigate into the directory (cd image-blaster). Install the Claude CLI using curl -fsSL https://claude.ai/install.sh | bash. Users must provide API keys for World Labs and FAL. Place your input image in the input/ directory and instruct Claude to "blast it," confirming each step. The process aims to deliver a 3D environment in under 5 minutes.

Highlighted Details

• Generates diverse 3D assets: textured/low-poly/geometry-only 3D models (.glb, .obj), Gaussian splat environments (.spz), and ambient/physics sound effects (.mp3).
• Assets are designed for integration into major game engines (Unity, Unreal, Godot) and DCC software (Blender, 3DS Max, Maya), as well as web applications (Three.js, Electron).
• Leverages specific AI models: marble-1.1 (environment), hunyuan-3d (objects), elevenlabs-sfx (audio), nano-banana/gpt-image-2 (image editing).
• Hunyuan 3D model generation supports configurable face counts (50k-1.5M), PBR material generation, and output types (Normal, LowPoly, Geometry).

Maintenance & Community

No specific details regarding maintainers, community channels (like Discord/Slack), or project roadmap were found in the provided README content.

Licensing & Compatibility

The provided README content does not specify a software license. Consequently, commercial use and compatibility with closed-source projects remain undetermined.

Limitations & Caveats

The project's development notes indicate that modifying the .claudeignore file is necessary to grant Claude permission to alter the React viewer, suggesting potential limitations or ongoing work on the viewer's customization. No other explicit limitations, known bugs, or alpha/beta status were mentioned.