Jump to content

AI Art Generation Handbook/How to install Stable Diffusion

From Wikibooks, open books for an open world

There are 2 main methods of using Stable Diffusion, one is installing it locally on the PC itself and others are using via web platforms.

For this chapter, we are focusing more to installing it locally on your computer with many added advantages such as privacy (as you won't need to save the generated images over clouds), able to tweak settings and installing extensions to enhance the image generation capability. However, you need a good graphics card so that it doesn't take long (e.g. minutes) to generate an image. This page describes how you can install it.

Hardware Requirements

[edit | edit source]

There are a few PC system pre-requisite that are needed to install Stable Diffusion on the PC

Minimum Recommended Specs Remarks
6GB VRAM

(Video RAM - GPU memory)

8GB VRAM*

(For SDXL model base)


12GB VRAM*

(For Flux Model)

Your machine are able to generate much larger images without Out of Memory (OOM) error message although the community are trying to optimize with lower VRAM but advance tweaking is required.
16GB RAM Your machine can handle the processing of multiple image generations in parallel
64GB Hard Disk Your machine can store more checkpoints/LORAS and generated images

Software Pre-Requisites

[edit | edit source]

Python 3.10.12 (Confirmed working as of July 2024)

Git Version Control

Web-UI (Recommended the most popular Automatic1111, but there are others too, see list of web-ui list below)

List of Local AI Art WebUI

[edit | edit source]

For a list of local (installed in the local PC) web-UI

List of local AI Art WebUI

Basic GitHub Command

[edit | edit source]

To install the WebUI into Local Hard disk, you need to know some of the GitHub Commands(1) git clone  : Clone the repository into the local hard disk

(2) git pull  : Update the local WebUI into the latest version

(3) git checkout : Revert the existing version back to earlier stable versions

Local AI Art WebUI Installation Guide

[edit | edit source]

Auto1111 (Recommended for ease of use , installation and extension compability)

ComfyUI (Recommended if you prefer node-based programming ala Blender)

Invoke.AI (Recommended if you like powerful in-painting functions)

SD.Next (Recommended if you prefer more support for other models such as Deepfloyd , Kadinsky and others)

Forge (Recommended for low memory VRAM management)

Sharing Models Between Different Local AI Art WebUI

[edit | edit source]

Optional but highly recommended:

All the Models , LORAs, etc.. are advisable to be kept in a central location in your Hard Disk (to save the precious spaces in your HD).

SD.Next

[edit | edit source]

Once you have successfully installed the SD.Next, go to System tab on main menu bar , then Settings at below of the main menu bar and scroll down on the left bar until you see System Paths. On text field with Folder with stable diffusion models , replaced it with the actual locations of where you stored the models.

Click Apply Settings buttons and the Base Model should appeared on top of the top left


References

[edit | edit source]

[1] https://medium.com/@promptingpixels/sharing-stable-diffusion-models-between-different-applications-856fc93c45e5