Skip to content

COMFYUI_AND_3D_GUIDE.md

Latest commit

 

History

History
68 lines (48 loc) · 3.29 KB

COMFYUI_AND_3D_GUIDE.md

File metadata and controls

68 lines (48 loc) · 3.29 KB
Raw

ComfyUI & 3D Mesh Generation Setup Guide

This guide explains how to set up ComfyUI with 3D Mesh Generation nodes (TRELLIS V2 and Hunyuan3D v2) and connect it to LocalLLMServerManager.

🚀 Overview

LocalLLMServerManager acts as a central proxy, VRAM orchestrator, and interactive WebGL 3D asset studio. It connects to your local ComfyUI instance running on http://127.0.0.1:8188.

Key Benefits

  • Automated VRAM Offloading: Automatically unloads Ollama LLM models from GPU memory before launching heavy ComfyUI 3D mesh or FLUX image generation workflows.
  • Interactive 3D WebGL Canvas: Renders generated .glb and .gltf 3D meshes natively in the browser with 360° orbital controls, wireframe toggles, and direct downloading.
  • Pre-Configured Workflow Presets: Ships with API-format workflow JSON templates for TRELLIS V2, Hunyuan3D v2, and FLUX / SDXL image generation.

🛠 Prerequisites & Installation

1. ComfyUI Setup

Ensure ComfyUI is installed on your local machine:

  • Download & install ComfyUI.
  • Launch ComfyUI using run_nvidia_gpu.bat (or your preferred launch script). ComfyUI defaults to http://127.0.0.1:8188.

2. Installing 3D Mesh Nodes

To run 3D generation workflows, install the relevant custom nodes via ComfyUI-Manager:

A. TRELLIS V2 Nodes

  • Search for and install ComfyUI-Trellis or TRELLIS custom nodes.
  • Place TRELLIS weights inside ComfyUI/models/trellis/ or models/checkpoints/.

B. Hunyuan3D v2 Nodes

  • Search for and install ComfyUI-Hunyuan3DWrapper or Hunyuan3D-v2.
  • Download Hunyuan3D v2 weights into ComfyUI/models/hunyuan3d/.

⚙️ Connecting to LocalLLMServerManager

  1. Open the manager dashboard at http://localhost:5246.
  2. Navigate to the 3D & ComfyUI Studio tab.
  3. Check the ComfyUI health badge in the top-right header:
    • 🟢 Online: ComfyUI is running and responding at 127.0.0.1:8188.
    • 🔴 Offline: ComfyUI is not running. Launch ComfyUI or update the endpoint URL in the configuration bar.
  4. (Optional) In the configuration bar, set your Preferred Image & Mesh Engine to ComfyUI or Stable Diffusion / Forge.

🎨 Running 3D Workflows

  1. Select your target preset:
    • TRELLIS V2 (3D Mesh Generator)
    • Hunyuan3D v2 (3D Mesh Generator)
    • FLUX / SDXL (High-Res Image Generator)
  2. Customize the prompt in the text box (e.g. a futuristic cyberpunk sports car, 3d asset, white background).
  3. Click 🚀 Queue Generation (Auto-Free LLM VRAM).
  4. The status box will show progress as the server automatically unloads loaded Ollama LLM models from GPU VRAM and posts the workflow payload to ComfyUI.
  5. Once complete, click any generated .glb asset in the Generated 3D Mesh Assets gallery to load it interactively in the WebGL 3D viewer!

📤 Custom Workflow Uploads

You can export custom workflows from ComfyUI for use in LocalLLMServerManager:

  1. In ComfyUI, enable "Enable Dev mode Options" in settings.
  2. Click "Save (API Format)" to export the workflow as a JSON file.
  3. Save the JSON file into the Workflows/ directory inside LocalLLMServerManager.
  4. The workflow will automatically appear in the preset dropdown menu!