Prompt for Midjourney

How Midjourney actually reads a prompt

Midjourney does not parse a prompt like a human reading a sentence; it converts your words into a set of weighted concepts and steers the image toward all of them at once. Two things follow from that. First, word order matters: terms near the front of the prompt carry more influence than terms near the end, so the subject and the two or three qualities you care about most should lead. Second, the model has no idea what you left out — if you do not name the lighting, the lens, or the mood, it fills those gaps with whatever its training data considers average. The prompt above fixes both problems by forcing a deliberate, front-loaded description covering subject, style, lighting, composition, mood, palette, and fine detail.

Using parameters instead of describing them

The biggest mistake newcomers make is trying to describe technical settings in plain English when Midjourney has a dedicated flag for them. Aspect ratio is set with --ar (for example --ar 16:9 for a widescreen frame or --ar 2:3 for a portrait), not by writing "wide image." The model version is set with --v 6, which changes how literally your prompt is interpreted — version 6 follows longer, more specific prompts far more faithfully than older versions. For photographic realism, add --style raw, which strips out Midjourney's default painterly "beautification" and gives you flatter, more true-to-life results. Other useful flags include --stylize (0 to 1000, controlling how much artistic license the model takes) and --chaos (0 to 100, controlling how varied the four initial grid images are).

Writing descriptors that actually change the picture

Vague adjectives like "beautiful" or "amazing" do almost nothing because they contain no visual information. Concrete descriptors do the work. Instead of "good lighting," specify "golden hour backlight" or "soft north-facing window light" or "hard noon sun with sharp shadows." Instead of "nice style," name a medium and reference: "35mm film photograph," "watercolor illustration," "isometric 3D render," or an art movement like "art deco." You can also weight individual terms with a double colon — forest::2 cabin::1 tells Midjourney the forest matters twice as much as the cabin. Composition cues such as "close-up," "wide establishing shot," "rule of thirds," or "shot from below" give you control over framing that the model would otherwise guess at.

Iterating toward the shot you want

A first render is a starting point, not a verdict. When something is close, use Midjourney's variation and reroll tools rather than rewriting from scratch, and change only one variable at a time — swap the lighting, or the style, or the aspect ratio, but not all three at once — so you can see what each change did. Keep the prompts that work in a personal library; the specific phrasing that reliably produces a look you like is genuinely reusable, and building that vocabulary is what separates consistent results from lucky ones.

Why this prompt works

Midjourney weights the front of the prompt most and responds strongly to parameters. This prompt orders descriptors by importance, appends the right flags for aspect ratio and version, and offers variations — the workflow that produces striking, controllable results.

How to customize it

• Put the most important descriptors first.
• Match --ar to where the image will be used.
• Try --style raw for more photographic, less stylized output.

Prompt variations to try

Photorealistic portrait

Write a Midjourney prompt for a portrait of [subject]. Comma-separated, most important first: the person and expression, camera and lens (e.g. 85mm f/1.4), lighting setup (e.g. Rembrandt lighting, softbox key), background, mood, and skin/texture detail. Append --ar 4:5 --v 6 --style raw. Use concrete photographic descriptors, no vague adjectives. Give 2 lighting variations.

Logo / flat graphic

Write a Midjourney prompt for a flat logo or icon of [subject]. Describe, comma-separated: the core shape, style (flat vector, minimal, geometric), color palette (2-3 colors), background (solid or white), and any symbolism. Append --ar 1:1 --v 6 and a high --stylize value if you want more flair. Keep it clean and simple. Give 2 palette variations.

Environment / concept art

Write a Midjourney prompt for an environment concept for [scene]. Comma-separated, ordered by importance: the location and time of day, art direction (e.g. Studio Ghibli, gritty sci-fi), atmosphere and weather, composition (wide establishing shot, foreground/midground/background layers), color palette, and mood. Append --ar 16:9 --v 6 --chaos 20 for varied grid options. Give 2 mood variations (one warm, one cold).

Common mistakes to avoid

• Describing technical settings in words instead of using flags. Writing "square widescreen" confuses the model; set the frame with --ar 1:1 or --ar 16:9 instead.
• Piling on empty adjectives like "beautiful, stunning, amazing." They add no visual information. Replace them with concrete cues — a named lighting style, a medium, a lens, a palette.
• Ignoring word order. Midjourney weights the front of the prompt more heavily, so burying your subject at the end weakens it. Lead with subject and the qualities you care about most.
• Forgetting --style raw for realism. Without it, version 6 applies a painterly polish that undercuts photographic looks. Add it whenever you want true-to-life output.
• Changing everything at once between renders. If you rewrite the subject, style, and lighting together, you cannot tell what caused the change. Adjust one variable per iteration.

Frequently asked questions

What is the difference between --v 6 and --style raw?

Do longer prompts always give better images?

How do I keep a consistent character or style across images?

Why do my results look nothing like the prompt?

Tip: replace the parts in [square brackets] with your own details before you send. The more specific you are — audience, tone, goal, constraints — the better the AI output.