Stable Diffusion 3.5 Definitive Setup Guide (ComfyUI / Forge)

Introduction

Stable Diffusion 3.5 (Large) is an incredibly powerful model, but it requires some environment setup to run. This guide introduces two methods: **"ComfyUI for Extensibility"** and **"WebUI Forge for Ease of Use."**

🤔 Which one should you choose?

ComfyUI (Method A): Recommended for those who want to understand the mechanics, try the latest techniques immediately, or maximize memory efficiency.
WebUI Forge (Method B): For those who prefer the familiar operation of the original A1111 and want to generate images easily via a "standard Web UI."

Pre-check

Common Prerequisites

✅ Python 3.10.x ～ 3.11.x (Forge is most stable on 3.10)
✅ Git
✅ VRAM 12GB+ (For comfortable use of SD3.5 Large)

⚠️ Note: SD3.5 Large is quite heavy. If you have 8GB VRAM or less, "Flux 2 (KLEIN)" might run more smoothly for you.

Method A: ComfyUI (Recommended & Portable)

Option 1

Manual Construction of ComfyUI

This is the cleanest and easiest way to manage your environment without cluttering your system.

                    # 1. Create folder & clone
                    mkdir C:\AI_Tools
                    cd C:\AI_Tools
                    git clone https://github.com/comfyanonymous/ComfyUI.git

                    # 2. Create virtual environment
                    cd ComfyUI
                    python -m venv venv

                    # 3. Install dependencies (CUDA version of PyTorch)
                    .\venv\Scripts\pip install torch torchvision torchaudio --index-url
                    https://download.pytorch.org/whl/cu121
                    .\venv\Scripts\pip install -r requirements.txt
                

Where to Place Your Models

Download sd3.5_large.safetensors from Hugging Face and place it as follow:

Main Model: ComfyUI/models/checkpoints/
CLIP (if needed): ComfyUI/models/clip/

Launch Command

.\venv\Scripts\python main.py

Method B: WebUI Forge (Browser-based UI)

Option 2

One-Click Setup for WebUI Forge

This is a high-speed fork of the former Automatic1111 and is currently the most active browser UI project.

                    # Just Git Clone and run
                    git clone https://github.com/lllyasviel/stable-diffusion-webui-forge.git
                    cd stable-diffusion-webui-forge

                    # Initial launch (Environment will be built automatically)
                    webui-user.bat
                

💡 Auto Installation: The initial launch will take some time as it downloads several gigabytes of data. Once finished, your browser will open automatically.

Key Settings

When using SD 3.5, enabling the --fp8 option in the "UI Settings" or as a launch argument will dramatically reduce VRAM usage.

🚑 Too much trouble to set up?!

With Antigravity, you just say, "Set up an SD 3.5 environment," and it's done. Leave dependency resolution and Git commands to the AI while you sit back and enjoy your coffee.

Read Antigravity Review

Stable Diffusion 3.5 Setup Guide

Introduction

🤔 Which one should you choose?

Common Prerequisites

Method A: ComfyUI (Recommended & Portable)

Manual Construction of ComfyUI

Where to Place Your Models

Launch Command

Method B: WebUI Forge (Browser-based UI)

One-Click Setup for WebUI Forge

Key Settings

🚑 Too much trouble to set up?!

🎨 Recommended Models for Comparison

Introduction

🤔 Which one should you choose?

Common Prerequisites

Method A: ComfyUI (Recommended & Portable)

Manual Construction of ComfyUI

Where to Place Your Models

Launch Command

Method B: WebUI Forge (Browser-based UI)

One-Click Setup for WebUI Forge

Key Settings

🚑 Too much trouble to set up?!

🎨 Recommended Models for Comparison

Flux 2 Setup Guide

Midjourney v7

Grok Image