## 🧱 Breaking Free from the Browser: Building a Standalone 3D Generation App with Hunyuan3D-2
**Hi, I’m Andre — an independent developer and seasoned computer enthusiast with a passion for creative problem-solving and human-friendly design.**
Over the years, I’ve learned that the most powerful tools are often hidden behind complex interfaces, online dependencies, or cloud-reliant APIs. As a developer who values local-first computing and seamless UX, I decided to change that — one app at a time.
Today, I’m excited to introduce something I’ve been building for months:
**C:\code:>_3D Generator** — a **fully standalone desktop application** for generating and texturing 3D models with Tencent’s powerful [Hunyuan3D-2](https://github.com/Tencent/Hunyuan3D-2). No Gradio, no browser tabs, and no constant internet connection required once setup is complete.
—
### 🛠 Why I Built This
Most public implementations of Hunyuan3D — including the official demo and Blender addon — either require a cloud backend or deep Blender integration. I wanted something that:
– ✅ Runs *entirely* offline after installation
– ✅ Has a clean, native desktop interface
– ✅ Supports **images and text prompts** for 3D generation
– ✅ Applies **AI-powered textures** and lets you retexture in real time
– ✅ Shows results with integrated 3D previews, thumbnails, and HDR lighting
This app is the result: a PySide6-powered GUI with embedded web-based mesh preview, background mesh optimization, and support for the full text-to-3D and image-to-3D pipeline — all *locally*.
—
### 🚀 Highlights
#### 🔹 Intuitive UI
– Drag and drop an image or type a prompt.
– Generate a shape, texture it with a single click.
– Real-time preview and thumbnail history.
#### 🖼 Smart Texturing
– Generate textures from text.
– Use your own image.
– Or apply variations from previously generated thumbnails.
#### 🌄 HDRI Support
– Drop in an environment map to enhance lighting instantly.
#### 🧠 Ollama-Powered Descriptions
– Auto-generate text prompts from image uploads using local LLMs like `moondream`.
#### 🔒 No Internet Needed (After Setup)
Once the models are downloaded, everything runs locally. Your data, your machine.
—
### 🧩 Behind the Scenes
The application glues together several powerful components:
– `Hunyuan3DDiTFlowMatchingPipeline` for generating meshes
– `Hunyuan3DPaintPipeline` for texturing
– PyMeshLab and Trimesh for mesh optimization
– PySide6 for the interface
– Ollama (optional) for local image captioning
– A local server and `QWebEngineView` to display `.glb` meshes
Everything is modular and cleanly split — and yes, **no Gradio**.
—
### 📦 How to Get Started
Check the [README](https://github.com/dahandla) for full instructions, including:
– Conda environment setup
– Dependencies
– Model downloads (once — then it’s offline!)
Once launched, you’ll never need to fire up Blender or wait for a web backend again.
—
### 🛍 Who It’s For
– Artists who want a faster pipeline from idea to mesh
– Developers looking to explore 3D-AI workflows locally
– Anyone who prefers working **offline**, without reliance on servers or fragile browser demos
—
### ❤️ Support & Contributions
This project is a labor of love. If you find it useful, consider:
– ⭐ Starring the repo
– ☕ Supporting via [PayPal](https://www.paypal.com/paypalme/AndreDickinson?locale=en_US)
– 🔁 Sharing with other AI & 3D enthusiasts
—
**Technology is both a tool and an art form.**
This project reflects my belief in making it accessible, reliable, and a little bit fun.
— Andre Dickinson
[github.com/dahandla](https://github.com/dahandla)