Stable Diffusion is an AI model for text to image generation. There are various GUIs available to tweak settings, add specific styles like Disney or photorealistic (Lora models) and help adjusting existing images via picture to picture or selection infill.

The guides below focus on the Automatic1111 Stable Diffusion WebUI.

This docker requires docker as well as the GPU drivers installed on the host system.

https://github.com/AbdBarho/stable-diffusion-webui-docker

git clone https://github.com/AbdBarho/stable-diffusion-webui-docker
cd stable-diffusion-webui-docker
vi docker-compose.yaml
#adjust name to stable-diffusion if desired
#adjust data and output dir location to /opt/stable-diffusion/data and output respectively if desired
mkdir -p /opt/stable-diffusion/{data,output}

#Build the ai and download the models.
#Note, first start will take 15-60 minutes to download models to data folder as cache depending on internet connection
#size approximately 10GB

docker compose --profile download up --build

# if download errors occur, just repeat the command.

# build the desired interface:
# docker compose --profile [ui] up --build
# where [ui] is one of: invoke | auto | auto-cpu | comfy | comfy-cpu

# use auto-cpu or comfy-cpu for cpu only interface

#docker compose --profile auto-cpu up --build
docker compose --profile auto up --build

# later:
docker compose --profile auto up -d

access the app on:

http://localhost:7860/

Windows AMD AUTOMATIC1111 stable diffusion webui using Microsoft DirectML for GPU acceleration:

1. Download and run installers for Python 3.10.6 (ticking Add to PATH) and git
2. Clone the fork as AMD is not officially supported
   Paste into command prompt in a directory where it should live
   

   git clone https://github.com/lshqqytiger/stable-diffusion-webui-directml
   cd stable-diffusion-webui-directml
   git submodule init
   git submodule update
   venv\Scripts\python.exe -m pip install --upgrade pip
   venv\Scripts\pip.exe pip install -r requirements.txt
   venv\Scripts\pip.exe install torch-directml

3. Right-click webui-user.bat and adjust the command line args to
   

   COMMANDLINE_ARGS=--use-directml --skip-torch-cuda-test

4. If you have only 4-6gb vram, try adding these additional flags to webui-user.bat's command line argslike so:
   

   --opt-sub-quad-attention --lowvram --disable-nan-check

5. Double-click webui-user.bat in that directory, this will download ~2.7GB torch and other python modules, followed by the 4GB Stable-Diffusion 1.5 pruned model.
6. If it looks like it is stuck when installing or running, press enter in the terminal and it should continue.
7. once done, a browser opens to http://localhost:7860/

Check/Test if GPU is used in Task Manager's GPU tab. Dedicated GPU Memory should show large usage for the loaded Model and when generating a test image in the web interface, the GPU usage should ramp up like here:

Approx times to generate 512×512 default settings images:

As an extension for SD-webui:
Open the Extensions tab in SD-webui.

1. Select the Install from URL option.
2. Enter https://github.com/zanllp/sd-webui-infinite-image-browsing
3. Click on the Install button.
4. Wait for the installation to complete and click on Apply and restart UI.

As an extension for SD-webui:
Open the Extensions tab in SD-webui.

1. Select the Install from URL option.
2. Enter https://github.com/Nevysha/Cozy-Nest
3. Click on the Install button.
4. Wait for the installation to complete and click on Apply and restart UI.

Some general Stable Diffusion prompt advice

More promptng techniques

Inpainting explanation

LoRA models/styles/tools as well as base models can be searched and downloaded from CivitAI

Description for photorealistic images:

1. Download model Realistic Vision V6.0 B1 Base Model from https://civitai.com/models/4201/realistic-vision-v11 and save it into the models/stable-diffusion directory
2. create models/Lora directory
3. download LORA Style hyperrealism art and save to models/Lora directory
4. download LORA epi_noiseoffset and save to models/Lora directory
5. in web interface go to extensions → install from url and paste to install:
   

   https://github.com/mcmonkeyprojects/sd-dynamic-thresholding
   https://github.com/opparco/stable-diffusion-webui-composable-lora

6. then go to installed tab and click apply and quit
7. restart stable diffusion web ui
8. go to Lora tab, configure entries (which will create json files, then the entries disappear)

Base settings:

set sampling steps to 20
set sampling method to DPM++ SDE Karras
set width to 768 and height to 512
set CFG scale to 6
set seed to -1
enable Dynamic Thresholding (CFG Scale Fix)
enable all composable Lora

cd /opt/stable-diffusion/data/models/Stable-diffusion
wget https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0/raw/main/sd_xl_base_1.0_0.9vae.safetensors

This is for general objects/scenes/animals/etc, faster than SDXL

download SDXL Turbo from https://huggingface.co/stabilityai/sdxl-turbo and copy into Models/Stable-diffusion directory.

https://civitai.com/models/176555?modelVersionId=214296

Other Text LoRAs:
https://civitai.com/tag/text

Lexica
PromptHero

3d, 8k, masterpiece, (best quality), Japanese garden, waterfall, cherry blossoms in full bloom, conifer, koi pond, (pagoda:1.1), stone lantern, water basin, water reflections, winding path, (night sky:1.:1), hokusai inspiration, ultra-realistic, pastel color scheme, soft lighting, golden hour, tranquil atmosphere, landscape orientation

EasyNegative, (worst quality:1.2), (low quality:1.2), (lowres:1.1), (monochrome:1.1), (greyscape), multiple views, comic, sketch, watermark

25 steps, 2m karras, 1024x768 seed 1192013237

greg rutkowski, highly detailed, dark, surreal scary swamp, terrifying, horror, poorly lit, trending on artstation, incredible composition, masterpiece

with settings:
 LMS Karras
 50 steps
 CFG scale 8
 random seed

Test. use in txt2img the following prompt:

RAW photo, close up photo of face, woman with dark hair in rain at night, blue eyes, detailed eyes, dark skin, (high detailed skin:1.2), 8k uhd, dslr, soft lighting, high quality, film grain, Fujifilm XT3   <lora:lora_contrast_fix:1>  <lora:lora-style-hyperrealism-art:1> 

and as negative prompt below:

(deformed iris, deformed pupils, semi-realistic, cgi, 3d, render, sketch, cartoon, drawing, anime:1.4), text, close up, cropped, out of frame, worst quality, low quality, jpeg artifacts, ugly, duplicate, morbid, mutilated, extra fingers, mutated hands, poorly drawn hands, poorly drawn face, mutation, deformed, blurry, dehydrated, bad anatomy, bad proportions, extra limbs, cloned face, disfigured, gross proportions, malformed limbs, missing arms, missing legs, extra arms, extra legs, fused fingers, too many fingers, long neck 

then click generate as often as you like

https://github.com/AbdBarho/stable-diffusion-webui-docker/issues/683

If you want a more up to date version of A1111 here are the steps:

docker exec -it webui-docker-auto-1 /bin/bash
git pull
pip install -r requirements.txt
#if needed
cd repositories
git clone https://github.com/AUTOMATIC1111/stable-diffusion-webui-assets.git
exit

docker stop webui-docker-auto-1
docker start webui-docker-auto-1

If it fails for any reason just rebuild the container using the standard process of docker compose –profile auto up –build