# image-blaster

> An open-source Claude skillset that converts a single image into a fully meshed 3D environment, 3D object models, and spatial sound effects in under 5 minutes.

image-blaster is an open-source TypeScript project by GitHub user neilsonnn that gives Claude the ability to transform a single input image into a complete 3D world — including explorable environments, object meshes, and ambient sound effects. Licensed under MIT, it is freely available on GitHub and designed to jumpstart 3D asset creation workflows for game developers, filmmakers, architects, and hobbyists alike.

## What It Is

image-blaster is an image-to-world skillset for Anthropic's Claude AI assistant. Rather than being a standalone application, it operates as a set of Claude skills that orchestrate multiple third-party generation APIs — World Labs, FAL (Hunyuan 3D), and ElevenLabs — to produce a full suite of 3D assets from a single photograph or concept image. The project's README describes the pipeline as capable of taking "an image to a fully meshed 3D environment in < 5 minutes."

## How the Pipeline Works

When a user drops an image into the `input/` directory and prompts Claude to "blast it," the tool executes a multi-step generation sequence by default:

- **3D object models** (`.glb`, `.obj`) are created for all *dynamic* objects in the scene using the Hunyuan 3D model via FAL.
- **Gaussian splat** (`.spz`) of the *static* environment is generated using World Labs' `marble-1.1` model.
- **Ambient and object-specific sound effects** (`.mp3`) are produced using ElevenLabs' SFX model.

Image editing and cleanup tasks use `nano-banana` by default, with `gpt-image-2` available as an alternate provider.

## Generation Models and Parameters

image-blaster integrates several specialized AI models under the hood:

- `marble-1.1` — World Labs' Marble model for explorable spatial environments
- `nano-banana` — default image editing for source cleanup and object reference images
- `gpt-image-2` — alternate image edit provider
- `hunyuan-3d` — Tencent's Hunyuan 3D model via FAL for object mesh generation
- `elevenlabs-sfx` — ElevenLabs sound effects for ambient and physics-based audio

Advanced users can tune Hunyuan 3D output with parameters including face count (40,000–1,500,000), PBR material generation, polygon type (Normal, LowPoly, Geometry), and triangle vs. quadrilateral polygon mode.

## Target Audience and Use Cases

The project's README lists a broad range of creative and professional use cases: video game level concepts, architectural renderings, film location scouting, robotics environment simulation, and personal/nostalgic scenes. It is explicitly designed to embed under the assets of game engines (Unity, Unreal, Godot), DCC software (Blender, 3DS Max, Maya), and web apps (Three.js, Electron).

## Setup Path

Getting started requires cloning the repository, entering the directory, and running Claude via its CLI installer. Users then provide API keys for World Labs and FAL directly to Claude in the terminal session. No separate configuration file setup is described beyond placing an input image in the `input/` folder and issuing a natural-language prompt.

## Current Status

The repository was created in April 2026 and last updated in May 2026, with the project's README describing it as actively maintained. The GitHub page reports 3,798 stars and 370 forks as of the last recorded update, indicating rapid early community interest. Five open issues are listed. The project is under active development, with a note in the README about an optional React viewer that can be unlocked by modifying the `.claudeignore` file.

## Features
- Convert a single image to a fully meshed 3D environment in under 5 minutes
- Generate 3D object models (.glb, .obj) for dynamic scene objects
- Create Gaussian splat (.spz) of static environments via World Labs marble-1.1
- Produce ambient and object-specific sound effects (.mp3) via ElevenLabs
- Image cleanup and editing via nano-banana or gpt-image-2
- Configurable Hunyuan 3D parameters: face count, PBR, polygon type
- Embeddable in Unity, Unreal, Godot, Blender, Three.js, and Electron apps
- Operates as a Claude skillset via natural-language prompts
- Optional React viewer for in-browser preview

## Integrations
Claude (Anthropic), World Labs (marble-1.1), FAL (Hunyuan 3D), ElevenLabs SFX, gpt-image-2 (OpenAI), nano-banana, Unity, Unreal Engine, Godot, Blender, 3DS Max, Maya, Three.js, Electron

## Platforms
WEB, API, CLI

## Pricing
Open Source

## Version
main

## Links
- Website: https://github.com/neilsonnn/image-blaster
- Documentation: https://github.com/neilsonnn/image-blaster#readme
- Repository: https://github.com/neilsonnn/image-blaster
- EveryDev.ai: https://www.everydev.ai/tools/image-blaster