Prompt engineering is a skill in directing a generative AI application to render what you would like, by using descriptive text and structured input. The process is similar to what a movie director has to do when imagining a scene.
There are four main components to a well-structured prompt:
1) The Main Subject. You can think of this as a character or cast in a play. You can be very descriptive about the appearance of the character(s), including things like eye and hair color, age, gender, wardrobe, and so on. Other attributes could include ethnicity, emotional state, height and weight, the era in which they lived, and more.
2) Secondary Elements. You can think of this as a scene or a setting, including props. You can be as descriptive as you like about the scene and what is in it, including attributes about the setting and the other elements such as their condition, proximity to each other and to the subject, scale, and many other attributes.
3) Lighting, Camera, and Mood. Providing details aboiut the scene such as where the camera is positioned, what time of day it is, and even the atmosphere can help to advance your narrative. For the camera alone there are many considerations, such as:
Close Up
Extreme Close Up
Medium Shot
Wide Angle
Fisheye
High or Low Angle
Motion Blur
Depth of Field
4) Style and Aesthetic. These attributes of the scene determine how the generative AI application will render the scene, not what it will render. Visual styles for example can leverage Art History to achieve a desired effect. Examples include:
Medium - Oil, Watercolor, Pen and Ink, and so on
Genre - Cubism, Expressionism,
Byzantine, Pop Art, Art Deco, and many more
Region - European, Affrican, Latin, etc.
Era - Renaisance, Mediaeval, 20th Century
Uploading a reference image will help to provide additional input for the application to compose the scene. You can set how closely the AI will match the reference image, from loosely to very strict.