Logo of Trellis2
Trellis 2
FeaturesPricing
BlogFAQ
Logo of Trellis2
Trellis 2

Transform images into stunning 3D models with AI-powered technology

Product

  • Features
  • Pricing
  • Blog
  • FAQ

Company

  • About
  • Contact

Legal

  • Privacy Policy
  • Terms of Service

© 2026 Trellis 2. All rights reserved.

TRELLIS 2 ComfyUI Workflow: Image to 3D GLB Step by Step (2026)
2026/04/21
6 min read

TRELLIS 2 ComfyUI Workflow: Image to 3D GLB Step by Step (2026)

Detailed step-by-step ComfyUI workflow for converting images to 3D GLB models using TRELLIS 2. Includes downloadable workflow JSON, node configuration, parameter explanations, and troubleshooting.

Last updated: April 21, 2026

This tutorial walks through a complete ComfyUI workflow that takes a single image and produces a production-ready 3D GLB model with PBR textures using Microsoft's TRELLIS 2. Every node, parameter, and connection is explained — so you can reproduce it exactly or modify it for your own needs.

If you haven't installed TRELLIS 2 in ComfyUI yet, start with our TRELLIS 2 ComfyUI installation guide.

Workflow Overview

The complete pipeline consists of 5 stages:

Input Image → Image Preprocessing → TRELLIS 3D Generation → Texture Baking → Export GLB

Required Nodes

StageNodeSource
InputLoad ImageComfyUI built-in
PreprocessingImage Resize / Center CropComfyUI built-in
3D GenerationTRELLIS Image to 3DCustom node
TextureTRELLIS Texture BakeCustom node
ExportTRELLIS ExportCustom node
PreviewPreview 3D / Save ImageComfyUI built-in

Hardware Requirements for This Workflow

ResolutionMinimum GPUVRAMApprox. Time
512³RTX 30608 GB45-60s
1024³RTX 407016 GB90-120s
1536³RTX 4090 / A10024 GB60-90s

Step 1: Prepare Your Input Image

TRELLIS 2 works best with well-prepared input images. Before loading into ComfyUI:

Ideal Input Characteristics

  • Resolution: 512x512 to 1024x1024 pixels
  • Background: Solid color or transparent PNG
  • Subject: Single object, centered, fully visible
  • Lighting: Even, diffused — avoid harsh shadows
  • Format: PNG (preferred) or JPG

What to Avoid

  • Multiple objects or cluttered scenes
  • Blurry or low-resolution images (below 256x256)
  • Heavy occlusion (parts hidden behind other objects)
  • Extreme perspective distortion

Step 2: Build the Workflow

Node 1: Load Image

Add a Load Image node to the canvas.

  • Drag and drop your prepared image, or click to browse
  • Supported formats: PNG, JPG, WebP

Node 2: Image Preprocessing (Optional but Recommended)

Add an Image Resize node to standardize the input:

Settings:

  • Width: 512
  • Height: 512
  • Resample Method: lanczos
  • Crop: center

This ensures consistent results regardless of your input image dimensions.

Connect: Load Image → IMAGE → Image Resize → IMAGE

Node 3: TRELLIS Image to 3D

This is the core generation node. Add TRELLIS Image to 3D to the canvas.

Configure these parameters:

ParameterSettingWhy
Resolution1024Good balance of quality and speed
Seed-1 (random)Use a fixed seed for reproducibility
Steps30Default works well; increase to 50 for fine details
CFG Scale7.5Controls adherence to input image
Shape Guidance0.5Lower = more creative, higher = more faithful
Low VRAMfalseSet to true if you have less than 16GB VRAM

Connect: Image Resize → IMAGE → TRELLIS Image to 3D → image

Node 4: TRELLIS Texture Bake

Add TRELLIS Texture Bake to apply PBR materials to the generated mesh.

Configure these parameters:

ParameterSettingWhy
Texture Resolution1024Sharp enough for most applications
PBR ChannelsfullGenerates Base Color + Roughness + Metallic + Normal
Bake MethodsmartBetter UV utilization
Denoise Strength0.6Smooths texture noise

Connect: TRELLIS Image to 3D → MESH → TRELLIS Texture Bake → mesh Connect: TRELLIS Image to 3D → IMAGE → TRELLIS Texture Bake → reference_image

Node 5: TRELLIS Export

Add TRELLIS Export to save the final model.

Configure these parameters:

ParameterSettingWhy
FormatglbBest for web, Unity, Unreal, Three.js
Mesh SimplifylowReduces polygon count without visible quality loss
Embed TexturestrueEverything in one file
Apply TransformtrueCenters and scales model appropriately

Connect: TRELLIS Texture Bake → MESH → TRELLIS Export → mesh

Node 6: Save Output (Optional Preview)

Add Save Image for a 2D preview of the textured result.

Connect: TRELLIS Texture Bake → PREVIEW → Save Image → images

Complete Node Connection Map

[Load Image]
     │
     ▼ IMAGE
[Image Resize] 512x512 lanczos
     │
     ▼ IMAGE
[TRELLIS Image to 3D] res=1024, steps=30
     │              │
     ▼ MESH         ▼ IMAGE
     │              │
     ▼              ▼
[TRELLIS Texture Bake] tex=1024, PBR=full
     │              │
     ▼ MESH         ▼ PREVIEW
     │              │
     ▼              ▼
[TRELLIS Export]   [Save Image]
  format=glb

Step 3: Run the Workflow

  1. Double-check connections — Ensure every node's output connects to the correct input
  2. Click Queue Prompt (or press Ctrl+Enter)
  3. Monitor progress — ComfyUI shows progress bars on each node
  4. Wait for completion — With 1024³ resolution on an RTX 4070, expect ~90 seconds total

Step 4: Inspect Your Output

Viewing the GLB File

  • Online: Upload to 3D Viewer or Autodesk Viewer (free, no signup)
  • Blender: File → Import → glTF 2.0
  • Windows 10/11: Native 3D Viewer (right-click → Open with → 3D Viewer)
  • Mac: Reality Converter or Xcode

Quality Checklist

After generating, check these aspects:

CheckWhat to Look For
Shape accuracyDoes the 3D model match the input image?
Surface smoothnessNo visible artifacts, holes, or jagged edges
Texture qualityColors match the input, no stretching or seams
TopologyClean mesh without non-manifold geometry
ScaleModel is centered and reasonably sized

Advanced Variations

Variation 1: Text-to-3D Pipeline

Replace the Load Image node with text input:

[CLIP Text Encode] "a futuristic helmet with neon lights"
         │
         ▼ CONDITIONING
[TRELLIS Text to 3D] res=1024
         │
         ▼ (same as above)

Variation 2: Multi-Image Pipeline

Use multiple angles for higher quality:

[Load Image 1] [Load Image 2] [Load Image 3]
      │              │              │
      └──────────────┼──────────────┘
                     ▼
         [TRELLIS Multi-View to 3D]
                     │
                     ▼ (same as above)

Variation 3: SDXL → TRELLIS Pipeline

Generate the input image with Stable Diffusion XL, then convert to 3D:

[SDXL Model Loader]
         │
         ▼
[CLIP Text Encode] "a detailed wooden chair"
         │
         ▼
[KSampler] steps=30, cfg=7
         │
         ▼ IMAGE
[TRELLIS Image to 3D]
         │
         ▼ (same as above)

This is the most powerful pipeline — you can generate any 3D model from pure imagination.

Performance Optimization Tips

Speed

  • Use 512³ for iteration, 1024³ for final output
  • Reduce steps from 30 to 20 for faster drafts
  • Disable texture baking during shape iteration
  • Use fp16 precision if your GPU supports it

Quality

  • Use 1536³ for production assets
  • Increase steps to 50 for complex subjects
  • Use multi-view input when possible
  • Enable full PBR baking for game engines

Memory

  • Close all browser tabs and GPU-heavy apps
  • Use Low VRAM mode on 8GB cards
  • Set batch size to 1
  • Process one model at a time

Common Workflow Errors

"Node connection type mismatch"

TRELLIS nodes use specific data types. Common mismatches:

  • IMAGE output → must connect to image input (not CONDITIONING)
  • MESH is a custom type — only connect between TRELLIS nodes

"Workflow runs but produces no output"

  • Check the model weights directory: ComfyUI/models/trellis/
  • Verify the TRELLIS custom node is in ComfyUI/custom_nodes/
  • Check ComfyUI console for error messages

"Generated 3D model is distorted"

  • Input image quality is the #1 factor
  • Try a different seed value
  • Increase resolution to 1024³
  • Ensure the subject fills most of the frame

Related Guides

  • TRELLIS 2 ComfyUI Installation Guide — How to set up TRELLIS 2 in ComfyUI
  • How to Use TRELLIS 2 — General usage guide
  • How to Install TRELLIS 2 — Standalone installation
  • TRELLIS 2 Download Guide — Direct model download links

FAQ

Can I download this workflow as JSON?

Community workflows are available on RunComfy and various Patreon creator pages. You can also save your own workflow from ComfyUI: Menu → Save.

How do I batch process multiple images?

Use ComfyUI's Queue system. Load multiple images via a directory node, or create an image list and connect it to a batch-aware TRELLIS node.

Can I automate this workflow?

Yes. ComfyUI provides an API endpoint (POST /prompt). You can script it with Python to process images in bulk.

What if I don't have a GPU?

Use the online version — no GPU needed. Upload your image and download the GLB file.

All Posts

Author

avatar for Trellis2 Team
Trellis2 Team

3D technology specialists focused on AI-powered 3D model generation, format conversion, and browser-based 3D rendering. We test and review 3D tools so you don't have to.

Categories

    Workflow OverviewRequired NodesHardware Requirements for This WorkflowStep 1: Prepare Your Input ImageIdeal Input CharacteristicsWhat to AvoidStep 2: Build the WorkflowNode 1: Load ImageNode 2: Image Preprocessing (Optional but Recommended)Node 3: TRELLIS Image to 3DNode 4: TRELLIS Texture BakeNode 5: TRELLIS ExportNode 6: Save Output (Optional Preview)Complete Node Connection MapStep 3: Run the WorkflowStep 4: Inspect Your OutputViewing the GLB FileQuality ChecklistAdvanced VariationsVariation 1: Text-to-3D PipelineVariation 2: Multi-Image PipelineVariation 3: SDXL → TRELLIS PipelinePerformance Optimization TipsSpeedQualityMemoryCommon Workflow Errors"Node connection type mismatch""Workflow runs but produces no output""Generated 3D model is distorted"Related GuidesFAQCan I download this workflow as JSON?How do I batch process multiple images?Can I automate this workflow?What if I don't have a GPU?

    More Posts

    How to Use TRELLIS 2 Online Free: No GPU Required (2026)

    How to Use TRELLIS 2 Online Free: No GPU Required (2026)

    Complete guide to using Microsoft TRELLIS 2 online for free. Covers Hugging Face Spaces, Google Colab, and other cloud platforms — no GPU or Python installation needed.

    avatar for Trellis2 Team
    Trellis2 Team
    2026/05/04
    Image to 3D Model for 3D Printing: Complete Workflow Guide (2026)

    Image to 3D Model for 3D Printing: Complete Workflow Guide (2026)

    Turn any photo into a 3D-printable model. Covers AI tool selection, mesh repair in Blender, STL export settings, slicer configuration, and troubleshooting common print failures. Works with TRELLIS 2, Meshy AI, and Tripo AI.

    avatar for Trellis2 Team
    Trellis2 Team
    2026/05/03
    TRELLIS 2 PBR Texture Generation in ComfyUI: Complete Guide (2026)

    TRELLIS 2 PBR Texture Generation in ComfyUI: Complete Guide (2026)

    How to generate production-quality PBR textures (Base Color, Roughness, Metallic, Normal maps) with TRELLIS 2 in ComfyUI. Covers texture baking nodes, parameter tuning, and export for Unity/Unreal.

    avatar for Trellis2 Team
    Trellis2 Team
    2026/04/21

    Newsletter

    Join the community

    Subscribe to our newsletter for the latest news and updates