-
-
Notifications
You must be signed in to change notification settings - Fork 135
Getting Started
Stable Diffusion is a text-to-image AI created by stability AI, like Dall-E and Midjourney it can be used to generate images from text!
It works by starting with an image filled with random noise and then denoises that to create an image.
if you want to go ahead and start generating images right away, feel free to jump to AI Horde section below.
this is probably the most used model of Stable Diffusion, most fine-tuned models use this as the base model.
some commonly used finetuned models and merges will be discussed in the Models page.
TODO: expand this section
Some of the main parameters of Stable Diffusion are explained below:
Prompt: this is the main prompt you use to tell the AI what to generate.
Negative Prompt: what you DON'T want in the image.
Steps: how many times you want to denoise the image.
CFG: how much closer should the image be to the prompt, lower values give more creativity to Stable Diffusion and higher values make the image closer to the prompt but may lead to noisy images.
Seed: A number that Stable Diffusion uses to generate the noise, if you want to make a small change without affecting the image too much you should use the same seed (note: some samplers such as those with _a (ancestral) at the end generate different images even with the same seed)
Highres. Fix: it's useful for generating larger images, if you've tried generating images of higher resolution you probably noticed there being multiple of the same thing often merged together, this is because SD was trained on images of size 512×512 and it doesn't generate bigger images well, that's where Highres fix comes in, by generating an image at a lower resolution and then using img2img to upscale it, the composition remains the same as that of the smaller image while the resolution is bigger.
AI Horde is a platform on which people can share their GPU and earn kudos in return (which give higher priority) more info in FAQ, it is completely free!
To get started, just pick a client from client interfaces page, visit/download it, and you can just start generating images.
Did you know we also have a discord that you can join and share your art, ask questions, get involved with the community, and earn kudos in?
There's also some unofficial servers
- https://discord.gg/ugEqPP5wMT (more NSFW oriented)
Don't forget to register for an account API at https://aihorde.net/register, then use the command /login on both servers above to be able to receive and gift kudos!
To get more priority, you need to register (it does NOT require an e-mail, nor any other PII.) and get an API key (don't forget to log in with the key in the client you're using), this will also enable you to make images of higher resolution.
You can also run your own worker and earn kudos!
Instructions for running your own worker can be found in AI Horde Worker repo. don't have a GPU powerful enough to run a worker ? don't worry, you can still contribute by using Google colab (notebook link), make sure to put the worker into maintenance before stopping.
you can also run an alchemist worker (upscaler/interrogation) which can run on lower spec devices.
This is a list of ways you can earn kudos:
- Rating images
- Sharing art in the Discord server
- Running a Worker
- Completing bounties on the official horde discord bounty forum.
- Supporting db0 on Patreon or Github (monthly kudos reward)
- Improving this wiki or
- Just asking for kudos in the Discord server
Stable Diffusion Art - A website with tutorials for Stable Diffusion.
Stable Diffusion cheatsheet
Stable Diffusion links hub (Rentry) - has tutorials too
Haidra Architecture - flow chart illustrating how AI Horde works