01Set up a repeatable first session
Begin on the Midjourney website with an account and the service access required by your plan. Locate the Imagine bar, settings, creation feed, and the controls for opening an individual result. Before generating, record the selected image model, aspect ratio, active personalization or references, and a small session limit. A saved prompt without its surrounding settings is an incomplete experiment. Use one fictional brief rather than exploring several unrelated ideas at once.
02Write an image brief before a prompt
A useful image brief states what must be visible, who will view it, how it will be used, and what would make it unsuitable. Translate that brief into a subject, action or pose, environment, framing, light, medium, and a short set of important constraints. Describe the visual result rather than asking the generator to understand a business goal such as increase sales. Keep uncertain creative choices separate from requirements that must survive every revision.
03Compose with framing and negative space
Composition directs attention through scale, placement, contrast, and empty space. Decide the main focal point before adding decorative details. A wide landscape can communicate environment while a close crop can communicate material. Aspect ratio changes the canvas shape, not merely the amount of blank border. Plan where later text or cropping will occur and inspect whether important details survive the intended placement.
04Direct light, material and visual medium
Lighting descriptions work best when they identify source direction, softness, color, and the surfaces that reveal them. Material words should support visible evidence: ceramic glaze, brushed metal, rough paper, or translucent glass. A photographic prompt and a flat graphic prompt ask for different kinds of detail. Change one lighting or medium decision at a time so a useful result can be explained rather than attributed to a long stack of aesthetic phrases.
05Use image parameters as controlled variables
Place supported image parameters after the descriptive text. Separate prompt wording from model and parameter choices in your record. A parameter comparison should ask one question, such as whether a wider aspect ratio serves the brief or whether more stylization changes subject clarity. Keep the current model explicit. Do not copy a legacy quality, reference, or speed option into a new model without checking compatibility.
06Use an AI assistant as a prompt editor
An assistant can reorganize a brief, find ambiguous words, suggest alternatives, and turn a visual critique into a revision plan. It cannot verify an image it has not seen or guarantee that a remembered Midjourney parameter remains supported. Give the assistant the actual brief, a short list of verified controls, and a request to preserve the key subject facts. Review its suggestions as drafts rather than commands that must be executed.