Hello all and welcome to RunDiffusion!
Stable Diffusion is a very large playground with tons of features and things to explore, however, if you are just starting up and have no idea where to begin, this is the place to be.
If you’re starting from scratch and prefer to learn through video, our free Stable Diffusion masterclass for beginners walks through foundational concepts and practical workflows at your own pace. Use it alongside this guide to understand how your prompts and settings work together.
We will begin our Stable diffusion journey in txt2img, this will be the default tab to start up on.
Before we begin prompting, we will want to select our model, I suggest to start with RDFX 2.5D in the model selection drop down.
Now that we've chosen our model, lets move onto our prompt.
Prompt: The prompt section is where you input all your ideas that you want to create into an image. You can find out more information on detailed prompting here: https://learn.rundiffusion.com/prompting-101/
For now, lets start with something simple like a "cat playing with a ball". Then click the orange "generate" button on the top right.
Perfect, a cat playing with a ball!
How about "A man walking in the woods"
Looking good!
Feel free to start adding descriptive details into the prompt. "A portrait of a large man walking through a forest with a red brick path"
Negative prompt: The negative prompt is telling StableDiffusion everything you DO NOT want in your prompt. I suggest with StableDiffusion V1.5 you start with no negative prompt and add in anything you must as you generate. Meanwhile, with StableDiffusion V2.1 it is very important to add negatives and mainly won't get very good images without a negative prompt. It is important to note there is no perfect negative prompt, it all has to do with what you are generating, and what ends up becoming visible in the image then adding descriptive negative prompts to combat it.
Here are a few suggestions: Blurry, low quality, distorted, bad artist.
Moving along to the settings of our images. These can be found directly under the two prompt windows.
This is where your image starts to come alive. Here you will find the following options.
Sampling Method: I recomend sticking with Euler A or DPM++ 2m karras, However more information can be found here: https://learn.rundiffusion.com/sampling-methods/
Steps: This is how many steps Stable Diffusion will take to generate an image. I suggest starting with 20-40 steps.
Height and width: Chosing your aspect ratio.
Batch count: How many separate batches of images will be created. If you select 2, Stable Diffusion will create a full image, then another full image for a total of 2.
Batch size: How many images Stable Diffusion will create in one generation.
CFG scale: CFG scale is how closly you want Stable Diffusion to pay attention to your prompt. The long and more detailed your prompt becomes, the higher you can raise this setting. I suggest starting around 5-8 cfg.
These are the very first steps for diffusing your creativity into reality!
Below, I have added some prompts to follow along with or get ideas from. Remember, you are asking a robot to create an image from your words, be as descriptive as possible!
Warrior dog, armor, holding sword, ready to fight
anime style, 1 girl, walking in a stream, water, highly detailed
racecar driving down a mountain, mountains, fast car, highly detailed
Feel free to use these prompts, along side adding your own details to help further depict the image.
Lastly, ensure to hop in our discord, check out our channels, share your art and ask questions!
Pick a model and start generating
These models are tuned for detail — open one and put this guide to work.
