An open source desktop app turns any photo into a 3D mesh on your own GPU, skipping the cloud generation queue, the freelance 3D artist brief, and the subscription.
Ever tried to get a usable 3D asset out of a marketing or e-commerce team on a deadline? You've got two options: queue up in a cloud image-to-3D generator and hope the mesh comes out clean, or brief a freelance 3D artist and sit through a couple of revision rounds before you have something you can actually drop into a product page or a game engine. lightningpixel/modly offers a third option: point a photo at your own GPU and skip the queue and the freelancer entirely.
Modly is an open source desktop app that turns a photo, or a text prompt, into a 3D mesh, running entirely on local hardware. It ships as an installer for Windows, Linux, and Apple Silicon Macs. No account, no subscription, no per-generation credit meter. It has real traction behind it too: 5.9k stars, 598 forks, and 15 contributors, with enough steady shipping to have already reached 12 tagged releases.
The workflow is node based. Drag an image in, run it through a Generate Mesh node, add the result to a scene, then smooth or decimate the mesh before export. That's roughly the same sequence of decisions a 3D artist makes by hand, just compressed into a few clicks. Behind the scenes it leans on a plug-in system. The latest release, v0.4.1 from July 16, supports official extensions wrapping models like Hunyuan3D 2 Mini, Hunyuan3D 2 Mini Turbo, TripoSG, and Trellis2, so you're not locked into one generation model, you swap the extension depending on whether you want speed or fidelity. There's also a CLI built specifically so a script or an agent can call a running Modly instance headlessly, kick off a generate job, poll it, and pull the exported mesh file without ever opening the UI.
This lands squarely with the people who currently pay for renders rather than the people who make them. A solo e-commerce founder who wants a hero product shot turned into a rotating 3D viewer. A marketing ops person mocking up a pitch deck. An indie game studio that needs placeholder assets without waiting on a contractor. An agency that wants to offer 3D previews without hiring a specialist. If your team owns even one GPU with enough memory to run a diffusion model, this replaces the "who do we ask to make this" conversation with "let's just run it."
None of that is free in the way "open source" sometimes implies. Someone still has to set up a Python environment, install the extension models, and troubleshoot when a workflow graph throws a validation error, that's an IT afternoon or a technically curious ops person's Saturday, not a five minute install. macOS support is Apple Silicon only. And local generation still trails a dedicated 3D artist on anything that needs to look genuinely polished, flagship hero assets are still going to a studio. Modly is MIT licensed, but with an attribution clause: fork it into your own product and you're required to credit "Based on Modly by Lightning Pixel" somewhere visible. That's a real obligation, not a suggestion, worth reading before anyone builds a business on top of it.
The interesting part here isn't that AI can turn a photo into a mesh, cloud tools have done that for a while. It's that the login, the queue, and the invoice just became optional.