HomeTutorials
Stable Diffusion Tutorial Navigation
46

Stable Diffusion Tutorial Navigation

B1Tee
B1Tee
July 26th, 2023

Elementary Tutorial

Local Deployment

Want to run stable-diffusion-webui and model smoothly , A minimum of 4GB video memory is required, the basic configuration is 6GB video memory, and 12GB video memory is recommended. Intel and AMD graphics cards are not recommended for deployment.

Mac platform deployment

Widows platform deployment(installer version)

Widows platform deployment

WebUI

Relationship between SD and WebUI

WebUI Instructions

WebUI Extended Function

Basic Model

The basic model determines the basic style of the generated screen, and it is also the basis for drawing. You can add character models such as LoRA on this basis to generate the most desirable pictures.

Common model styles & installation and use

Introduction and download of 12 basic large models

Model recommendation: anime style OrangeMixs

Due to the low training cost of small models, players are more inclined to train small models of various styles, you can find them at Websites such as station C download various small models.

Take LoRA as an example, download, install and use the small model If you want to train the model yourself, you can find tutorials in the advanced techniques section.

Keyword (prompt) writing

prompt syntax details

prompt to generate website

prompt to write Skills

Advanced Skills

WebUI Plugin: ControlNet Full Tutorial

One of the most powerful plugins at present, including image pose adjustment, recoloring, background replacement, prompt saving and many other functions.

VAE concept explanation

VAE introduction and usage instructions

Pose Adjustment

Using ControlNet or PoseX to Adjust Image Action/Pose

Model training

The "model training" here refers to the "small model", which is compared to the checkpoint "big model" with several GB on the network. The training time of these small models is short, the file is only about a few MB, and the training cost is not high.

LoRA model training

Textual Inversion (Embedding) model

HyperNetwork model

3 models Different training tendencies: - LoRA is more suitable for AI to learn the overall style of pictures. - Embedding, suitable for AI to learn a new concept/object, or the drawing of a single character. - HyperNetworks is more suitable for AI to learn the overall style of pictures.

Comments

no dataCoffee time! Feel free to comment