What it does Modly is a TypeScript desktop app for Windows and Linux that turns a single photo into a 3D mesh using open-source AI models. Everything runs locally on your GPU; no cloud upload, no subscription gate. The project ships with a Python backend and an Electron-style frontend that you can either install from a release or run from source with npm and a venv.

The interesting bit The extension system is the real architecture bet. Models aren’t baked into the binary; they’re GitHub repos with a manifest.json and a generator.py that Modly pulls in on demand. The project already maintains five official extensions wrapping Hunyuan3D, TripoSG, and Trellis2 variants, so the core team is essentially curating a plugin marketplace for generative geometry.

Key highlights

• Runs entirely local: models execute on your own GPU, no data leaves the machine
• Extension-based model loading: install new generators from GitHub URLs inside the app
• Five official extensions available at launch, including quantized (GGUF) and “turbo” speed variants
• Cross-platform desktop shell: Windows and Linux supported, macOS listed as coming soon
• MIT licensed with an explicit fork attribution requirement visible in the UI or docs

Caveats

• macOS support is promised but not yet shipped
• Setup from source requires juggling npm, Python venv, and pip dependencies—not quite “download and double-click”
• No benchmarks, hardware requirements, or model VRAM estimates are listed in the README

Verdict Worth a look if you need offline image-to-3D for privacy or cost reasons and don’t mind managing Python environments. Skip it if you’re on macOS or expecting a polished, one-click consumer experience; this is still a developer-grade tool with a GUI.