Modly is a free, open-source desktop app that converts images into 3D meshes using local AI models on your NVIDIA GPU.
It runs on Windows, Linux, and Apple Silicon macOS, and exports models as GLB, OBJ, STL, or PLY without ever uploading your images to an external server.
Drop a photo of a product, prop, or character concept into the app, and the app removes the background, runs a neural 3D reconstruction model on your GPU, and outputs a previewable mesh you can export directly to Unity, Blender, or your slicer.
The whole workflow stays on your machine, from the moment you load an image until the final export lands in your project folder.
Features
- Generates 3D meshes from a single image through local AI inference.
- Removes image backgrounds before mesh reconstruction.
- Exports generated models as GLB, OBJ, STL, and PLY files.
- Stores generated models in workspace collections.
- Smooths and decimates imported meshes inside the app.
- Installs model and process extensions from GitHub URLs.
- Includes Hunyuan3D 2 Mini quality, mesh-resolution, guidance, and seed controls.
- Runs local automation through a Python CLI with JSON output.
Use Cases
- Create early game props from concept art, then export GLB or OBJ files for a 3D package.
- Turn a product photo into rough reference geometry for a scene, mockup, or presentation.
- Generate an STL base mesh from a photographed object before mesh repair and slicer preparation.
- Build rough forms for Blender or ZBrush, then retopologize and refine surface detail.
- Run repeatable image-to-GLB jobs from a script or agent while the desktop app is active.
Modly vs Meshy & CSM
| Area | Modly | Meshy & CSM |
|---|---|---|
| Processing location | Your computer. | Provider servers. |
| Account | No account required. | Required. |
| Usage cost | No generation credits. | $20–60 per month |
| Privacy path | Images stay in the local workflow. | Images enter the provider’s processing path. |
| Hardware | Your GPU and system memory determine capacity. | Provider hardware handles generation. |
| Setup | Desktop installation, model downloads, and local storage. | Browser access and service onboarding. |
| Model control | Extensions can add model paths and workflows. | Model choices depend on the service. |
| Automation | Local CLI and JSON-based workflow runs. | API availability varies by provider. |
How to Use Modly
Basic Usage
1. Download the latest Modly app from the release page.
2. Open the Models page and select Install from GitHub. Paste the HTTPS URL for a supported extension, then confirm the installation.
3. Use a PNG, JPG, or WebP image with a clear subject and a visible silhouette. A single object against a simple background produces a more usable starting mesh than a crowded scene.
Modly removes the background automatically. Reflections, motion blur, hidden edges, and overlapping objects can still confuse reconstruction.
4. Set the generation controls. The recommended Hunyuan3D 2 Mini extension exposes these main controls:
| Control | Available values | What changes |
|---|---|---|
| Quality | Fast, Balanced, High | Diffusion steps: 10, 30, or 50. |
| Mesh Resolution | Low, Medium, High | Octree resolution: 256, 380, or 512. |
| Guidance Scale | 1.0 to 10.0 | Closeness to the reference image. |
| Seed | -1 to 4294967295 | Reproducible or random generations. |
Start with Balanced quality and Medium resolution. Increase the settings after the first mesh has the correct general shape. Higher values consume more VRAM and take longer.
5. Generate the mesh, then inspect the front, sides, and hidden areas in the built-in viewer. A single image cannot reveal the full back side of an object, internal geometry, or obscured details.
Use the transform controls to reposition the model. The app can also smooth and decimate imported meshes before writing the optimized result back to the workspace.
6. Export GLB for game engines and portable 3D scenes. Export OBJ for Blender and similar 3D tools. Export STL for 3D-printing workflows after checking scale, wall thickness, and mesh integrity. Export PLY for point-cloud and research-oriented workflows.
Model Extensions and Output Paths
Modly uses installable extensions instead of locking image-to-3D generation to one model.
| Extension route | Main role | Notes |
|---|---|---|
| Hunyuan3D 2 Mini | Single-image mesh generation | Includes Fast, Balanced, and High quality presets plus optional PBR texture support. |
| Hunyuan3D 2 Mini Turbo | Faster Hunyuan3D variant | Available through a separate extension package. |
| Hunyuan3D 2 Mini Fast | Lightweight Hunyuan3D variant | Available through a separate extension package. |
| TripoSG | Single-image 3D reconstruction | Listed as a separate Modly extension. |
| Trellis.2 GGUF | Geometry and texture workflow | Generates a GLB mesh, then supports a second texture pass. |
The Trellis.2 GGUF extension adds more detailed controls for image-to-mesh workflows:
| Control | Details |
|---|---|
| Pipeline | Fast, Balanced, High, or Ultra quality presets. |
| Quantization | GGUF quant levels from Q4_K_M through Q8_0. |
| Sparse structure steps | Controls the early structural diffusion stage. |
| Shape SLaT steps | Controls shape generation detail. |
| Foreground ratio | Controls how much of the frame the subject occupies. |
| Remesh resolution | Sets dual-contouring output resolution. |
| Texture resolution | Sets the final texture map size. |
| Atlas size | Sets UV atlas packing resolution. |
| Texture steps | Controls texture diffusion passes. |
| Seed | Repeats a selected generation path. |
Model weights and extension dependencies must download before the first generation. Each model can impose different memory requirements, quality tradeoffs, and license terms.
Modly CLI and Local Automation
Modly includes a Python CLI for scripts and agents. The desktop app must be running since the CLI communicates with a local API at 127.0.0.1:8765.
The CLI prints final machine-readable JSON to standard output. The --progress flag sends progress updates to standard error.
Check desktop readiness
python tools/modly-cli/agent.py healthUse compact one-line JSON for another tool or agent.
python tools/modly-cli/agent.py --compact healthInspect installed models
python tools/modly-cli/agent.py model list
python tools/modly-cli/agent.py model status
python tools/modly-cli/agent.py model params --model activeStart, inspect, and cancel workflow runs
python tools/modly-cli/agent.py workflow-run start --image ./input.png --wait
python tools/modly-cli/agent.py workflow-run status <run_id>
python tools/modly-cli/agent.py workflow-run cancel <run_id>Generate and export a mesh
python tools/modly-cli/agent.py generate \
--image ./input.png \
--output ./export.glb \
--progressUse --no-export when the workspace path is enough.
python tools/modly-cli/agent.py generate \
--image ./input.png \
--no-exportInspect capabilities and process runs
python tools/modly-cli/agent.py capability list
python tools/modly-cli/agent.py process-run statusLegacy compatibility commands
python tools/modly-cli/agent.py legacy job
python tools/modly-cli/agent.py legacy cancel
python tools/modly-cli/agent.py legacy generate --image ./input.png --output ./legacy.glbThe legacy commands target older /generate/* job endpoints. The current automation contract uses health, model, workflow-run, capability, and process-run.
Experimental workflow commands
python tools/modly-cli/agent.py experimental generate-from-workflow \
--workflow <name> \
--output <path>The experimental workflow command writes the final artifact to the path passed through --output. Image-only ComfyUI workflows continue through Modly’s image-to-3D path.
Run Modly From Source
The source route requires Node.js, Python, and the project dependencies.
Install JavaScript dependencies.
npm installCreate the Python environment.
cd api
python -m venv .venvActivate the environment on Windows.
.venv\Scripts\activate
pip install -r requirements.txtActivate the environment on Linux or macOS.
source .venv/bin/activate
pip install -r requirements.txtStart the development build.
npm run devRun the available tests and build checks.
npm test
./node_modules/.bin/tsc --noEmit -p tsconfig.node.json
npm run buildPackage the Apple Silicon macOS build.
npm run package:macRun the repository launch scripts when you need to start the project directly.
# Windows
launch.bat# Linux or macOS
./launcher.shAlternatives and Related Tools
- TripoSplat AI: Free Open-Source Image-to-3D Gaussian Generator
- CADAM AI: Free AI Text-to-3D and Image-to-3D Model Generator
- TRELLIS: Image and Text to 3D Asset Model
- Free AI Image Generation and Editing Tools
Pros
- Free local generations.
- No account required.
- Private image processing.
- Four export formats.
- Open-source MIT code.
- Local CLI automation.
Cons
- Local setup required.
- GPU-dependent workflow.
- Single-view geometry limits.
- No browser workspace.
