Installing and Testing Modly, a Local AI Image-to-3D Mesh Generation Tool

Why Choose Modly?

In game development and 3D design workflows, turning a 2D concept sketch into a 3D mesh model has always been a very time-consuming and technically demanding job. A number of cloud-based AI 3D generation tools have appeared on the market, but most of them use subscription pricing. They usually come with strict generation credit limits, and they also require you to upload your original designs to cloud servers. For commercial projects or privacy-sensitive creators, that is not the best fit.

Modly is a local AI 3D mesh generation tool open-sourced by Lightning Pixel. Its biggest feature is that it runs entirely on your local GPU, which means all your data stays 100% private, and you do not need to pay for cloud servers or a subscription. Modly provides a friendly desktop interface and supports a modular AI extension system, so you can easily load recent open-source 3D generation models such as Hunyuan3D 2 or Trellis.

Live Demo

Below is a live demo of Modly taking a single 2D image, generating a 3D mesh model in one click, and previewing it with real-time rotation:

How to Install and Get Started

Modly is straightforward to install. It supports ready-to-use installer downloads as well as a source-code startup flow for developers.

1. Download the Standalone Installer (Windows / Linux)

If you are using Windows or Linux, you can go directly to the Releases page and download the prebuilt installer:

Windows: download and run Modly-Setup-0.3.6.exe
Linux: download and run Modly-0.3.6.AppImage

2. Install and Run from Source (Windows / Linux / macOS Supported)

If you want to do local secondary development, or if you are using macOS, you can also clone the project repository and manually configure the Python and JS environments:

A. Clone the Project Repository

bash

git clone https://github.com/lightningpixel/modly.git
cd modly

B. Install Frontend Dependencies

bash

npm install

C. Set Up the Python Backend Environment

bash

cd api
python -m venv .venv

# Activate the Python virtual environment
.venv\Scripts\activate     # Windows
source .venv/bin/activate  # Linux / macOS

# Install the dependencies required by the model
pip install -r requirements.txt

D. Start the Development Server

bash

# Run this after returning to the project root directory
npm run dev

Or directly double-click/run the launcher script in the root directory:

Windows: run launcher.bat
Linux / macOS: run ./launcher.sh

Flexible Extension System

The main appeal of Modly is its highly flexible extension system. Each model extension is an independent GitHub repository containing manifest.json and generator.py, which lets the community package newly released 3D models and plug them into Modly very quickly.

Officially Recommended Supported Model Extensions:

Extension Name	Supported Model	GitHub URL
`modly-hunyuan3d-mini-extension`	Hunyuan3D 2 Mini	GitHub Repository
`modly-hunyuan3d-mini-turbo-extension`	Hunyuan3D 2 Mini Turbo	GitHub Repository
`modly-hunyuan3d-mini-fast-extension`	Hunyuan3D 2 Mini Fast	GitHub Repository
`modly-triposg-extension`	TripoSG	GitHub Repository
`modly-trellis2-gguf-extension`	Trellis2 GGUF	GitHub Repository

How to Install an Extension Module (Required Before Generation)

In the app, switch to the extension page and click Install from GitHub in the upper right.
Enter the HTTPS URL of the extension module you want to install, such as one of the links in the list above, and confirm.
After the module finishes installing, download the corresponding model weights or variants. Once that is done, you can start using it from the main screen.

Personal Notes

While using Modly for quick 3D modeling, what stood out most to me was its generation speed and workflow experience.

My test platform today was Windows 11 with an RTX 3070 Ti GPU. In actual testing, “image to 3D” generation with GLB export took only about 20 seconds, which was genuinely impressive. The installation process was also very simple, unlike many open-source AI projects that require a lot of complicated manual compilation.

Another nice part is that it offers a workflow similar to node-based tools. For users already familiar with Stable Diffusion or ComfyUI, the learning curve is almost nonexistent.

Small suggestion: Current 3D generation models, especially Trellis, have fairly high VRAM requirements. If you run into insufficient VRAM or OOM errors during generation, I recommend first trying the Mini Fast or Turbo extension versions, or lowering the resolution and quality settings in the generation configuration.

Related Links:

The software project introduced in this article is open-sourced under the MIT License. If you fork or further develop this project, make sure to follow the license terms and retain attribution to the original creator, Lightning Pixel.