text-to-3d-asset skill
Claude Code skill that turns text or a photo into a game 3D asset: Fooocus, Hunyuan3D-2, Blender and import into Unreal Engine.
- 148 ★ on the repository
- Last commit
Extended access
Why this level
- The scripts start and stop local servers, call Blender and write files, and the absolute paths inside need manual editing.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/text-to-3d-assetDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The scripts start and stop local servers, call Blender and write files, and the absolute paths inside need manual editing.
This is third-party code. Review the repository files before installing.
What it does
The `text-to-3d-asset` skill runs a chain of local AI services. Fooocus through Fooocus-API draws an image from text, Hunyuan3D-2 builds a textured `.glb` from it, and headless Blender converts the file to `.fbx` with embedded textures. The last step imports the result into an Unreal Engine project as a StaticMesh and Blueprint through the `unreal-mcp` server. The first and last stages are optional: with a photo in hand the chain starts at step two. `scripts/` holds three scripts: `fooocus_gen.py`, `hunyuan_gen.py` and `glb_to_fbx.py`. The author validated the chain on Windows with an RTX 4070 Laptop GPU with 8 GB.
Who it is for. Unreal Engine game developers with an NVIDIA GPU who want draft assets without manual modeling.
Good fit when
- A game needs a quick prop prototype: a locomotive, a bus, scenery.
- A photo of an object should become a textured model.
- A finished `.glb` has to be converted to `.fbx` for the engine.
- The model is imported into an Unreal project through MCP right after generation.
Not a fit when
- There is no NVIDIA GPU: Hunyuan3D-2 with textures is built for one.
- Work happens on macOS or Linux: the chain was validated only on Windows, and scripts hold paths like `C:\AI\`.
- A cloud service is more convenient than a local one: the chain needs Fooocus, Hunyuan3D-2 and Blender installed.
Example request
Make a 3D asset of a red double-decker bus from text and prepare an `.fbx` for Unreal import.Limitations
The author validated the chain on one Windows workstation with an NVIDIA RTX 4070 Laptop with 8 GB. `SKILL.md` and the scripts contain absolute paths like `C:\AI\` that must be changed for your install. On 8 GB GPUs Fooocus and Hunyuan3D must not be loaded together: generation that takes a minute balloons to about fourteen, so the skill stops the other server before running. Hunyuan3D-2 needs compiled `custom_rasterizer` and `differentiable_renderer` extensions matching the CUDA version. The last commit in the repository dates from July 2026.
How to disable. Delete the `text-to-3d-asset` folder from the project's `.claude/skills/` or from `~/.claude/skills/`. Stop the local Fooocus-API and Hunyuan3D servers separately.
FAQ
How do I install the skill?
Copy the folder into the project's `.claude/skills/` or `~/.claude/skills/`, then ask for, say, a red double-decker bus.
Can I start from a photo?
Yes, the first stage is skipped and the chain starts with Hunyuan3D-2, which builds a textured `.glb` from the picture.
Why stop a server before running?
Fooocus and Hunyuan3D together exhaust the memory of an 8 GB GPU, and generation stretches from one minute to fourteen. The skill stops Hunyuan (port 8080) before stage one and Fooocus-API (port 8888) before stage two.
Is Unreal Engine required?
No, import is optional. It needs the `unreal-mcp` server and an Unreal project.
Which files come out?
A `.png` image, a textured `.glb` model and an `.fbx` from Blender with embedded textures.
Discussion
No questions or reviews yet.
Related
SkillsA skill rebuilds the object in a reference image as a procedural Three.js model written in code, not downloaded art
A Claude Code skill that builds a game-ready 3D asset in Blender from concept art and checks each phase by measurement.
A Codex skill that builds a walkable Unreal Engine project from photos of a room or house, one question at a time.
A skill that turns narration text into a 20-second chunky-pixel explainer at 1920×1080 with voice-over and an editable project.
A Claude Code workflow that builds a dressed, lit Blender scene from reference images or one paragraph, with graded renders.
Thirty chained skills drive Blender through MCP: modeling, materials, lighting, camera, rendering, animation and glTF export
README in short
The skill organizes four stages: image, model, conversion and import. With a photo already at hand it starts at stage two, and Unreal is needed only for the last step.