Stable Diffusion Prompter Reference Guide

Understand that Stable Diffusion is a deep learning model designed to generate images from text descriptions, also usable for inpainting, outpainting, and image-to-image translation.

Create detailed prompts including keyword categories such as subject, medium, style, artist, websites, resolution, color, and lighting to improve image quality and specificity.

Incorporate popular keywords like 'digital painting,' 'portrait,' 'concept art,' 'hyperrealistic,' or 'pop-art' and reference specific artists or websites (e.g., Artstation, DeviantArt) to influence style.

Use syntax to adjust keyword importance, such as (keyword: factor), parentheses (), or curly braces {{}} to increase or decrease weight in the AUTOMATIC1111 GUI.

Apply keyword blending or prompt scheduling via keyword1: keyword2: factor syntax to transition between styles smoothly during generation.

Keep token limits in mind (typically 75 tokens per prompt in base models), and note that association effects (e.g., ethnic features with eye color or specific artist influences) can affect image output.

Iterate your prompt by starting simple and adding keywords gradually. Use universal negative prompts to avoid unwanted elements in the final image.

Recognize that early diffusion steps determine the main composition, and changing keywords later affects smaller details.