SDExp's Stable Diffusion Starter Guide
申し訳ありませんが、私は日本語が上手ではないので、代わりにこのウェブリンクガイドを提供します。
Work in Progress 仕掛かり品
Good news! There are many ways to use Stable Diffusion to generate anime art!
This guide will show various ways to setup Stable Diffusion from the simplest to the most complex.
As a beginner, start with Stable Horde to test things out.
Stable Horde (Great for beginners)
Stable Horde
Fast setup, free and simple to use. Slow generation for anonymous and new accounts. Start here to test things out.
Runs remotely from volunteer machines over the internet, so it can even run on your phone. Limited options to expand.
Stable Horde is a great way for beginners to get comfortable using stable diffusion and the basics of prompting. After using it, you can decide if you would like to delve deeper!
To speed up the processing time and generate larger images, you will need a free api key (no login needed, but it will speed up processing a bit more).
The image generation can take a minute or more for free accounts. You can speed this up drastically by rating images. I don't recommend using this long term, just to learn prompting basics and get a feel for stable diffusion.
- Get an API key here, then paste it here
- Click here to access Stable Horde with a test heart prompt.
- Click the blue "Generate" button to unlock the size settings. Wait for it to finish the image.
- Click one of my example images below to make it yourself!
- Click the blue "Generate" button and wait for it to finish.
- Edit the prompt to whatever you want to see!
- To see the full size image, click on the "images" tab on the top
You will get timed out for using certain prompts
To speed up image generation, rate images using the star icon on the left.
To make your own image from scratch, change the following settings to my recommended settings, and type these starter prompts (green setting only for API users):
Type what you want to see in the end of the "Prompt" box
If you want to go further, and have a powerful PC, I recommend installing Stable Diffusion locally on your PC via WebUI or Naifu
If you want to go further, and DO NOT have a powerful PC, I recommend running stable diffusion via Google Colab. Follow this guide.
Google Colab (Slow PCs)
Google Colab
Runs remotely from Google's machines over the internet, so it can even run on your phone.
Follow this guide.
Naifu (Windows)
Adapted from NovelAI leaks.
Simple to setup and install locally. A free local version of NovelAI.
Naifu is less flexible and slightly slower, so it is not recommended, unless you are having trouble with the setup of WebUI.
Minimum Requirements
Operating System = Windows 7 or newer
System Storage = 15GB
System RAM = 16GB
GPU = Nvidia Maxwell (GTX 7xx) or newer
GPU VRAM = 8GB
* If you are unsure about any of these specs, you can use diagnostic software like GPU-Z or Speccy.
- Download this torrent (magnet link)
- Install Python, checking "Add Python to PATH".
- Extract it like this
- Run setup.bat and wait for it to finish.
- Done!
To run Naifu:
- Run "run.bat" or "run.sh"
- Open http://localhost:6969/ in your browser.
- (Optional) Mobile usage: open http://computername:6969, http://computername.local:6969, or http://IPADDRESS:6969
WebUI (Recommended)
WebUI Stable Diffusion client
Most complicated to setup, but enables a large degree of flexibility and scalability.
Minimum Requirements
Operating System = Windows 7 or newer
System Storage = 20GB
System RAM = 16GB
GPU = Nvidia Maxwell (GTX 7xx) or newer
GPU VRAM = 2GB
* If you are unsure about any of these specs, you can use diagnostic software like GPU-Z or Speccy.
- Download the following files from this torrent (magnet link)
- Click here, and follow the guide.
That will get you a nice stable diffusion client with the novelai model.
Once you are more comfortable, feel free to experiment with different models, extensions, embeddings, vaes, LoRAs, etc.
Great resources for these: civitai.com & huggingface.co
I recommend these extensions:
Extension | Link |
---|---|
a1111-sd-webui-tagcomplete | https://github.com/DominikDoom/a1111-sd-webui-tagcomplete |
novelai-2-local-prompt | https://github.com/animerl/novelai-2-local-prompt |
sd-webui-additional-networks | https://github.com/kohya-ss/sd-webui-additional-networks |
stable-diffusion-webui-images-browser | |
stable-diffusion-webui-wd14-tagger | https://github.com/toriato/stable-diffusion-webui-wd14-tagger.git |
Notes On My Workflow
If you are familiar with Stable Diffusion, here is my workflow:
The model I use is a custom blend of novelai/anythingv3/protogen and 7th Layer. If you are a plus subscriber you can grab the original images and drag it into the "png2info" tab in the stable diffusion webui client, to see the exact prompt and parameters I used to make that image.
I use many custom embeddings, loras, vaes, and dreambooth models.
After generation, I use img2img and photoshop to finalize images.
As long as you are a fan, I can answer any questions you might have!