3D Mesh Reconstruction

The 3D Mesh Studio reconstructs textured 3D models from 2D images or text descriptions. It communicates directly with ComfyUI on port 8188 to run neural 3D reconstruction nodes.

The studio outputs production-ready 3D assets in industry-standard .glb and .obj formats.

---

Supported 3D Models

The studio integrates two modern open-source 3D reconstruction engines:

Engine	Primary Architecture	Input Modalities	Best Use Case
TRELLIS V2	Sparse-structure latents & multi-view flow matching	Image-to-3D & Text-to-3D	Sharp mechanical geometry, detailed props, and clean topology.
Hunyuan3D v2	Two-stage diffusion & textured mesh generator	Image-to-3D & Text-to-3D	Watertight organic models, creature sculpts, and complex characters.

NOTE

Install ComfyUI 3D custom nodes via ComfyUI-Manager. For detailed installation steps, refer to the ComfyUI & 3D Guide.

---

Hardware Requirements and Memory Allocation

Reconstructing 3D neural fields requires substantial GPU memory:

• Recommended VRAM: 12 GB to 16 GB dedicated GPU memory.
• Minimum VRAM: 8 GB (requires lower poly-count settings).
• Automated VRAM Offloading: The VRAM Orchestrator automatically clears Ollama language models from GPU memory before 3D reconstruction begins.

IMPORTANT

If your GPU has 8 GB of VRAM, close all other GPU applications before launching 3D workflows.

---

Generation Pipelines

The studio supports two distinct input pipelines:

1. Image-to-3D Pipeline

The Image-to-3D pipeline reconstructs geometry from a reference image:

• Supply an image with a single subject against a neutral background.
• The backend node automatically segments the subject and removes background noise.
• The pipeline estimates depth, multi-view consistency, and back-face geometry.

TIP

Generate your source image in the Image Studio using FLUX or SDXL. Include the phrase "isolated on white background, 3D asset" in your prompt.

2. Text-to-3D Pipeline

The Text-to-3D pipeline generates a multi-view visual concept first. It then converts that concept into an orbital 3D mesh automatically.

---

Step-by-Step 3D Mesh Generation

Follow these steps to reconstruct a 3D asset:

Step 1: Verify 3D Engine Status

1. Open the Studio tab in the left activity rail or navigation bar.
2. Click 🧊 3D Mesh in the top modality selector.
3. Verify the ComfyUI engine status card in the telemetry ribbon or header. Ensure the card indicates Online. If stopped, click ▷ Start directly on the ComfyUI card.

Step 2: Choose Generation Mode

Select your input mode:

• Text to 3D: Describe your desired object, materials, and form factors directly in the dock textarea.
• Image to 3D: Supply a reference image using the 📎 attachment button in the prompt dock or drag and drop a clean image into the canvas.

Step 3: Configure Geometry and Export Options

1. 3D Prompt: Enter descriptive details covering geometry, texture, and materials.
2. Settings Popover: Click ⚙️ Settings in the dock toolbar to configure generation parameters:
   • Export Format: Select GLB (PBR textures embedded) or OBJ (with .mtl).
   • Mesh Quality / Poly Count:
     • Standard: Balanced poly count optimized for real-time game engines.
     • High Detail: Dense mesh topology optimized for offline rendering and digital sculpting.

Step 4: Generate 3D Mesh

1. Click Generate ↵ in the bottom Creative Prompt Dock (or press Ctrl+Enter).
2. Monitor the 4-stage pipeline progress tracker on canvas.
3. Wait for the sparse-structure and texture generation passes to finish.
4. Inspect and interact with the rendered asset in the 360° orbital WebGL canvas.

---

360° Orbital WebGL Viewport

The right panel features an interactive WebGL canvas that renders generated .glb meshes in real time:

Viewport Navigation Controls

Interact with your 3D model using standard orbital camera controls:

• Rotate (Orbit): Click and drag with the left mouse button to rotate the camera around the asset.
• Pan: Click and drag with the right mouse button to shift the camera position horizontally and vertically.
• Zoom: Scroll the mouse wheel up or down to inspect fine texture details.
• Wireframe Toggle: Click the wireframe button to inspect polygon edge flow and quad density.
• Lighting Modes: Switch between studio lighting, flat ambient light, and untextured clay modes.

---

Exporting and Downloading 3D Files

Once reconstruction completes, click the export buttons below the 3D canvas:

• ⬇️ Export GLB:
  • Downloads a binary glTF file with embedded UV textures, normal maps, and PBR materials.
  • Compatible with Godot, Unreal Engine 5, Unity, and Three.js web applications.
• ⬇️ Export OBJ:
  • Downloads a standard wavefront OBJ file accompanied by MTL material definitions.
  • Compatible with Blender, Autodesk Maya, 3ds Max, and 3D printing slicers.

---

Custom ComfyUI 3D Workflows

You can export custom 3D workflows from ComfyUI and run them through Local LLM Server Manager:

1. Open ComfyUI in your web browser (http://127.0.0.1:8188).
2. Open the ComfyUI settings dialog and enable Enable Dev mode Options.
3. Click Save (API Format) to export your workflow as a .json file.
4. Copy the JSON file into the Workflows/ directory in Local LLM Server Manager.
5. Open the studio. The new custom 3D workflow appears automatically in the preset dropdown.

---

Related Documentation

• Studio Overview — Architectural overview and feature pack management.
• Image Generation Guide — Create reference images for Image-to-3D pipelines.
• ComfyUI Engine Setup — Install TRELLIS and Hunyuan3D custom nodes.