Everything Stable Diffusion

• Basic StableDiffusion
• Stable Diffusion
• Different versions of StableDiffusion
• Installation: Stable Diffusion Web UI by Automatic1111 (+ Deforum Extension)
• Installation: Deforum Stable Diffusion
• Helpful Links

Basic StableDiffusion

Stable Diffusion

Stable Diffusion version 2.1

stable diffusion v-2-1 768x768

wget --no-check-certificate --content-disposition https://th-koeln.sciebo.de/s/m3PhfQtlmYOlEUI/download -O v2-1_768-ema-pruned.ckpt

stable diffusion v-2-1 512x512 (base)

wget --no-check-certificate --content-disposition https://th-koeln.sciebo.de/s/Kpgz0TueJYcjY4t/download -O v2-1_512-ema-pruned.ckpt

MIDAS (Depth prediction)

wget --no-check-certificate --content-disposition https://th-koeln.sciebo.de/s/qzm306xfHdTm9Jk/download -O dpt_large-midas-2f21e586.pt

NYU AdaBins

wget --no-check-certificate --content-disposition https://th-koeln.sciebo.de/s/fTpi3vOXrNWhvK1/download -O AdaBins_nyu.pt

Stable Diffusion with depth2image

stable diffusion depth ema

wget --no-check-certificate --content-disposition https://th-koeln.sciebo.de/s/cOU4j8ZXa65Pmql/download -O 512-depth-ema.ckpt

dpt hybrid MidaS 

- should be placed in a folder named midas_models

wget --no-check-certificate --content-disposition https://th-koeln.sciebo.de/s/Ju8NlHuNDMFQ8SP/download -O dpt_hybrid-midas-501f0c75.pt

Stable Diffusion v1

🌟ProtoGen 3.4 🌟

wget --no-check-certificate --content-disposition https://th-koeln.sciebo.de/s/2MTmmPTeuSu20MY/download -O ProtoGen_X3.4.ckpt

Stable Diffusion v 1.5 pruned emanoly

wget --no-check-certificate --content-disposition https://th-koeln.sciebo.de/s/LwN49UHfYkVH2zE/download -O v1-5-pruned-emaonly.ckpt

Upscaling

x4-upscaler-ema.ckpt

wget --no-check-certificate --content-disposition https://th-koeln.sciebo.de/s/QmvMgK31iSQ4MS1/download -O x4-upscaler-ema.ckpt

Different versions of StableDiffusion

Installation: Stable Diffusion Web UI by Automatic1111 (+ Deforum Extension)

WebUI: Installation

1. Open a terminal

2. Run the following commands in terminal in this order:

Download the stable-diffusion-webui repository

git clone https://github.com/AUTOMATIC1111/stable-diffusion-webui.git

Change directory and navigate to the Stable Diffusion folder

cd stable-diffusion-webui/models/Stable-diffusion

3. Download a Stable Diffusion Model from KISD Model Zoo

Downloading ONE version is enough to get you started.

Stable Diffusion v-2-1 512x512 (base)

wget --no-check-certificate --content-disposition https://th-koeln.sciebo.de/s/Kpgz0TueJYcjY4t/download -O v2-1_512-ema-pruned.ckpt

Stable Diffusion v-2-1 768x768

wget --no-check-certificate --content-disposition https://th-koeln.sciebo.de/s/m3PhfQtlmYOlEUI/download -O v2-1_768-ema-pruned.ckpt

4. Change access rights of file webui.sh

Return back to home directory 

cd

Change directory and navigate to the stable-diffusion-webui folder

cd stable-diffusion-webui

Run this command to assign execution rights

chmod +x webui.sh

Optional: Install Deforum Extension

5. Run webui.sh

./webui.sh --share

6. In the terminal code, search for the public URL (Running on public URL: https:// xxxxxxx-xxxx-xxxx-.gradio.live) and copy it to your browser

7. If you receive the following error message after trying to generate your first image, in the WebUI go to Settings > Stable Diffusion > & check the box for "Upcast cross attention layer to float" > Apply settings & Reload UI

NansException: A tensor with all NaNs was produced in Unet. This could be either because there's not enough precision to represent the picture, or because your video card does not support half type. Try setting the "Upcast cross attention layer to float32" option in Settings > Stable Diffusion or using the --no-half commandline argument to fix this. Use --disable-nan-check commandline argument to disable this check.

8. At end of use: Within the interface, go to File > Hub Control Panel > Stop My Server

Deforum Extention (Animation for SD): Installation

1. Open a terminal & navigate to the stable-diffusion-webui folder

cd stable-diffusion-webui

2. Download the Deforum repository

git clone https://github.com/deforum-art/deforum-for-automatic1111-webui/ extensions/deforum

Tutorial recommendation: Deforum Settings Explained - Part 1 Stable Diffusion Automatic 1111

Tutorial recommendation: Deforum for AUTOMATIC1111 - Stable Diffusion Tutorial - AI Animation Part 2

After Installation: Running Stable Diffusion WebUI

1. Open a terminal & navigate to the stable-diffusion-webui folder

cd stable-diffusion-webui

2. Run webui.sh 

./webui.sh --share

3. In the terminal code, search for the public URL (Running on public URL: https:// xxxxxxx-xxxx-xxxx-.gradio.live) and copy it to your browser

4. At end of use: Within the interface, go to File > Hub Control Panel > Stop My Server

Tutorial recommendation: Stable diffusion tutorial. ULTIMATE guide - everything you need to know!

Please Note: The original version of this book was created by Laura Wagner. This is merely an update of her work!

Installation: Deforum Stable Diffusion

Installation

Installing DSD locally  on a Linux Device

1. clone this repository:

git clone https://github.com/HelixNGC7293/DeforumStableDiffusionLocal.git

Creating the environment

2. in case the environment should be available permanently (yes)

conda config --add envs_dirs /home/jovyan/.conda_envs

3. create environment with conda

conda create --name dsd python=3.8.5 -y

4. initialize bash shell

conda init bash

5. restart shell

source ~/.bashrc

6. activate environment:

conda activate dsd

Downloading models

cd into models folder

cd models

Download model files

wget --no-check-certificate --content-disposition https://th-koeln.sciebo.de/s/QJJm0HoP5JqMYza/download -O sd-v1-4.ckpt
wget --no-check-certificate https://th-koeln.sciebo.de/s/e9GlUImXPvTtJRP/download -O dpt_large-midas-2f21e586.pt

cd ..

cd pretrained

wget --no-check-certificate --content-disposition https://th-koeln.sciebo.de/s/yYEhjskz9yuKotX/download -O AdaBins_nyu.pt

cd ..

Making the installations

cd DeforumStableDiffusionLocal

python setup.py

Prompting the Model (Latest Stable Diffusion Weights)

These steps need to be executed in order to activate the environment we installed before

initialize bash shell

conda init bash

restart shell

source ~/.bashrc

then:

conda activate dsd

generate still images:

python run.py --settings "./examples/runSettings_StillImages.txt"

animation:

python run.py --enable_animation_mode --settings "./examples/runSettings_Animation.txt"

installation script

git clone https://github.com/HelixNGC7293/DeforumStableDiffusionLocal.git
conda config --add envs_dirs /home/jovyan/.conda_envs
conda create --name dsd python=3.8.5 -y
cd DeforumStableDiffusionLocal
cd models
wget ##############insert adabins
wget --no-check-certificate --content-disposition https://th-koeln.sciebo.de/s/QJJm0HoP5JqMYza/download -O sd-v1-4.ckpt
wget --no-check-certificate https://th-koeln.sciebo.de/s/e9GlUImXPvTtJRP/download -O dpt_large-midas-2f21e586.pt
cd ..
conda init bash
source ~/.bashrc
conda activate dsd
python setup.py

Downloading entire folders

if you want to download an entire folder you need to zip it first:

zip -r example.zip original_folder

Upscaling

For upscaling please refer to the ESRGAN Book

https://github.com/xinntao/Real-ESRGAN

-i INPUTPATH

-o OUTPUTPATH

-n MODEL

Helpful Links

a collection of different useful links

Prompt search engine - find good prompts

prompt exploration

openAI Microscope -explore CLIP

Stable Diffusion Examples on Twitter