How to Write Midjourney Prompts That Actually Work

Why prompt structure matters in Midjourney

Midjourney does not read a prompt the way a person reads a sentence. It weighs the words and phrases you give it, front-loads the ones that appear first, and tries to satisfy as many of them as possible at once. That means two prompts with the same words in a different order can produce very different images. Learning a repeatable structure is the single fastest way to go from random results to images that match what you pictured in your head.

The good news is that a strong Midjourney prompt almost always follows the same skeleton. Once you internalize it, you can describe anything — a product shot, a fantasy landscape, a logo concept — by filling in the same slots in the same order.

The anatomy of a great prompt

A reliable Midjourney prompt is built from seven parts, written in this order:

1. Subject — what the image is of (a red vintage motorcycle, a young woman reading).
2. Descriptive details — materials, clothing, age, expression, surroundings.
3. Style or medium — photograph, oil painting, 3D render, anime, concept art.
4. Lighting — golden hour, soft studio light, dramatic rim lighting.
5. Mood and color — moody and cinematic, warm pastel palette, high contrast.
6. Composition — close-up portrait, wide establishing shot, top-down flat lay.
7. Technical parameters — the flags such as --ar, --v, and --stylize.

Here is the skeleton applied to a real prompt: a red vintage motorcycle parked on a wet city street, chrome details, neon reflections, cinematic photography, dramatic night lighting, moody and atmospheric, wide-angle composition --ar 16:9 --v 6 --stylize 250. Every word earns its place, and the model has a clear hierarchy to follow.

Choosing the right style words

Style words are where most beginners either under-specify or pile on contradictions. Pick one primary medium — photograph, illustration, render — and then add one or two refinements. For photographic looks, naming a real lens or film stock (for example shot on 35mm, shallow depth of field) pushes the model toward realism. For illustration, naming an art movement or technique (watercolor, flat vector, chiaroscuro) is more effective than naming a living artist.

Avoid stacking five competing styles. Anime oil painting photorealistic 3D render forces Midjourney to average them into mush. Commit to a direction and let the other slots — lighting, mood, composition — do the differentiating work.

Understanding the key parameters

• --ar sets the aspect ratio (--ar 16:9 for widescreen, --ar 2:3 for portraits, --ar 1:1 for square).
• --v selects the model version. Newer versions render more coherent detail.
• --stylize (0–1000) controls how much Midjourney applies its own aesthetic. Low values stay literal to your words; high values look more artistic but drift from the prompt.
• --chaos (0–100) increases variety between the four initial results — useful when exploring.
• --no removes elements (--no text, watermark).

Start with just --ar and --v. Add --stylize only once your wording is solid, otherwise you will not know whether a change came from your words or the parameter.

A fast workflow: start from an image

The quickest way to write a strong prompt is to stop starting from nothing. If you already have a reference you like — a screenshot, a photo, a piece of concept art — you can reverse-engineer it into a prompt. Our free Image to Prompt Generator analyzes the colors, lighting, orientation and subject of any image directly in your browser and writes a Midjourney-formatted prompt (with the --ar and --v flags already filled in).

From there you edit: tighten the subject in your own words, swap the style if you want a different medium, and adjust the parameters. This image-first approach typically gets you to a usable result in two or three iterations instead of ten.

Example prompts you can adapt

Use these as templates and swap in your own subject and style:

• Product shot: a matte black wireless headphone on a marble surface, soft studio lighting, minimalist, neutral color palette, centered product photography --ar 1:1 --v 6
• Character portrait: portrait of an elderly fisherman, weathered face, knit sweater, soft window light, warm tones, shallow depth of field, photorealistic --ar 2:3 --v 6
• Landscape: misty pine forest at dawn, volumetric god rays, cool blue and green palette, wide cinematic composition, highly detailed --ar 16:9 --stylize 300
• Illustration: a cozy bookshop interior, flat vector illustration, warm pastel palette, soft ambient light, isometric composition --ar 4:3 --v 6

Notice how each one fills the same slots — subject, details, style, lighting, color, composition, parameters — just with different values. That consistency is what makes results predictable.

Common mistakes to avoid

• Burying the subject. If the most important thing appears at the end of a long prompt, Midjourney may under-weight it. Lead with it.
• Vague adjectives. Beautiful and amazing tell the model nothing. Soft pastel palette does.
• Contradictory instructions. Minimalist, highly detailed, busy background cannot all be true.
• Forgetting the aspect ratio. A portrait subject in a square frame wastes half your composition.

Iterating: variations, remix and weights

Your first four results are a starting point, not a verdict. When one is close, use Midjourney's variation buttons to explore nearby options, or upscale the best candidate. If you want to push composition harder, you can give two ideas different importance using multi-prompts and weights — for example cyberpunk city::2 rain::1 tells Midjourney the city matters twice as much as the rain.

Keep a short text file of prompts that worked. Over time you build a personal library of reliable phrasings for lighting, mood and style that you can paste into any new subject. Combined with starting from a reference image, this turns prompt writing from guesswork into a quick, repeatable process.