ERNIE-Image

ERNIE-Image ComfyUI Complete Workflow Setup: From Zero to Production Pipeline

mei 5, 2026

ERNIE-Image ComfyUI Complete Workflow Setup: From Zero to Production Pipeline

Step-by-step guide to building ERNIE-Image workflows in ComfyUI — from environment setup to advanced nodes.

Environment Setup

System Requirements

Component Minimum Recommended
GPU 12GB VRAM 24GB VRAM (RTX 4090)
RAM 32GB 64GB
Storage 50GB available SSD 200GB+
OS Ubuntu 22.04 / Windows 11 Linux preferred

Install ComfyUI

git clone https://github.com/comfyanonymous/ComfyUI.git
cd ComfyUI

python -m venv venv

source venv/bin/activate


pip install -r requirements.txt

Model Download and Configuration

Model File Structure

ComfyUI/
├── models/
│   ├── checkpoints/
│   │   └── ERNIE-Image/
│   │       ├── ernie-image-8b.safetensors
│   │       └── ernie-image-turbo.safetensors
│   ├── vae/
│   │   └── sdxl-vae-fp16-fix.safetensors
│   ├── clip/
│   │   └── clip-vit-large-patch14/
│   └── ipadapter/
│       └── ip-adapter-ernie.bin

Basic Workflow Setup

Node Connections

[CheckPointLoader] → ERNIE-Image 8B
    ↓
[CLIPTextEncode] → Prompt
    ↓
[EmptyLatentImage] → 1024x1024
    ↓
[KSampler] → Generate
    ↓
[VAEDecode] → Decode
    ↓
[SaveImage] → Output

Core Parameter Configuration

Node Parameter Recommended
KSampler steps 28 (Standard) / 8 (Turbo)
KSampler cfg 7.0
KSampler sampler euler_ancestral
KSampler scheduler normal
EmptyLatentImage width 1024
EmptyLatentImage height 1024

Advanced Workflow: IP-Adapter + ControlNet

Node Connections

[CheckPointLoader] → ERNIE-Image
    ↓
[CLIPVisionLoader] → CLIP-ViT-L-14
    ↓
[IPAdapterModelLoader] → IP-Adapter
    ↓
[IPAdapterApply] → Scale: 0.8
    ↓
[ControlNetLoader] → ControlNet model
    ↓
[ControlNetApply] → Depth Map
    ↓
[KSampler] → Generate
    ↓
[VAEDecode] → Output

Batch Generation Workflow

Node Connections

[CSVLoader] → Prompt list
    ↓
[BatchPrompt] → Batch processing
    ↓
[KSampler] → Batch generation
    ↓
[VAEDecode] → Batch decode
    ↓
[SaveImage] → Batch save

CSV Template

prompt,negative_prompt,seed,width,height
"a cat on a table","","42",1024,1024
"a dog in a garden","","137",1024,1024
"a bird in a tree","","256",1024,1024

Workflow Export and Sharing

Export JSON

import requests, json

workflow = {

"3": {"class_type": "KSampler", "inputs": {"steps": 28, "cfg": 7.0}},

}


with open("workflow.json", "w") as f:

json.dump(workflow, f, indent=2)

Load Workflow

response = requests.post(
    "http://localhost:8188/prompt",
    json={"workflow": workflow}
)

Common Issues

Q1: ComfyUI fails to start

Solution: Check Python version (3.10-3.11) and CUDA drivers.

Q2: Not enough VRAM

Solution:

  1. Lower resolution
  2. Use GGUF quantized model
  3. Enable --lowvram flag

Q3: IP-Adapter node not found

Solution: Install ComfyUI-Impact-Pack plugin.

Summary

ComfyUI workflow setup key steps:

  1. Environment: Python + CUDA + ComfyUI
  2. Models: ERNIE-Image 8B + VAE + CLIP
  3. Basic pipeline: Checkpoint → Prompt → KSampler → VAE → Save
  4. Advanced: IP-Adapter + ControlNet + Batch
  5. Production: CSV-driven + API calls

From zero to production-ready workflow, ERNIE-Image fully integrates with ComfyUI ecosystem.

This workflow uses ComfyUI + ERNIE-Image 8B.

ERNIE-Image Team