Image Generation

Stable Diffusion Prompting Guide

Learn how to craft effective prompts for Stable Diffusion, a deep learning model that generates images from text descriptions. This guide covers prompt structure, keyword categories, importance weighting, keyword blending, token limits, and iterative refinement strategies to prod

9 steps 1 variables English

Prompt template

Run these steps in order.

01
Understand that Stable Diffusion is a deep learning model designed to generate images from text descriptions, also supporting inpainting, outpainting, and image-to-image translation.
02
Construct detailed prompts by including keyword categories such as subject, medium, style, artist, website, resolution, color, and lighting to specify image output clearly.
03
Incorporate popular keywords like "digital painting," "portrait," "concept art," "hyperrealistic," and "pop-art," and reference artists or websites like Artstation and DeviantArt to influence artistic style.
04
Adjust keyword importance using syntax in tools like AUTOMATIC1111 GUI: (keyword: factor), parentheses (), or double curly braces {{}} to emphasize or de-emphasize aspects.
05
Use keyword blending or prompt scheduling with syntax such as keyword1: keyword2: factor to smoothly transition between styles during image generation.
06
Be aware of token limits (around 75 tokens per prompt in basic models) and consider association effects where attributes influence other image aspects like ethnicity or pose.
07
Iteratively refine prompts starting simple with subject, medium, and style, then add keywords progressively to enhance detail and quality.
08
Apply universal negative prompts to exclude undesired elements from generated images.
09
Remember early diffusion steps set overall composition while later keyword changes affect finer details and adjustments.

Prompt library

Use these prompts directly inside ChatGPT.

Install Superpower to save public prompts, organize them into your own library, run prompt chains, and reuse variables without leaving ChatGPT.