Remember, the best way to learn is still through practice and experimentation!

✍️ Prompt

A description of your image - more detailed descriptions will result in more detailed and accurate images.

Negative prompt

An argument that instructs the model not to include certain things in the generated image.


A model defines, in simple terms, the style of the resulting image.

📈 Steps

How many times to improve the generated image iteratively; higher values take longer; very low values can produce bad results. Notice how, in the example below, the image overall doesn't change too much anymore after only 15 steps. The recommended value to this parameter sits around 20 to 30 steps, any more than this and you start getting into diminishing returns territory.

🖌️ CFG Scale

Classifier Free Guidance Scale - how strongly the image should conform to the prompt - lower values produce more creative results - values too low might destroy coherency. Keep this value above 5 and below 15, recommended 7, if you aren't sure what value to use.

🔢 Seed

A seed controls the output of a random number generator. The same seed (and parameters) will generate the same image. Not specifying a seed, or rerolling, randomizes the seed.

Last updated