Modly setup: offline 3D generation and GLB export

This overview explains Modly, an open source desktop app for generating 3D meshes on your own hardware. The speaker describes Windows and macOS standalone installers and states that the app requires either an Nvidia GPU with at least 8 GB of VRAM or an Apple Silicon Mac. Its Electron interface includes a canvas and a live RAM usage panel.

The base app starts without a generation model. The walkthrough covers installing an extension through a supported GitHub URL, with Hunyuan 3D Mini and Trellis 2 as examples. According to the speaker, Modly creates isolated Python environments to separate PyTorch and CUDA dependencies. Model weights can require several gigabytes of downloads, so installation precedes the offline generation workflow.

The tutorial connects an image input node to a mesh generation node and a scene viewer. It describes validation warnings for incompatible connections before execution. After running the graph on the local GPU, users inspect the mesh and export a .glb file. The speaker gives a 10-to-15-second generation estimate, but does not tie that estimate to a specific model or hardware configuration.

The closing section discusses importing the export into Unreal Engine and adding PyMeshLab polygon reduction or auto-rigging extensions. The speaker calls the exported asset production ready; the overview does not establish that every generated mesh meets a project's production requirements.