How to Write Midjourney Prompts That Actually Work

July 4, 2026 · 6 min read · 1,555 words · Updated: July 2026

What you will learn

Most prompts fail for 3 reasons - missing layers beyond the subject, vague adjectives, and skipped visual language
The 6-layer framework - concept, subject, colors, composition, lighting, camera - turns vague prompts into specific ones
Replace vague words like "blue dress" with specific language like "deep cerulean silk midi dress with cool purple undertones"
See a direct before-and-after comparison of a prompt with and without the framework applied
Learn which Midjourney parameters matter most and the common mistakes to avoid

In this guide

Why most prompts fail
The 6-layer prompt framework
Putting it all together
Without and with the framework
The quick test
The Midjourney parameters that matter
Common mistakes to avoid
Try it now

Most people write Midjourney prompts like search queries.

'a woman in a red dress outside'

Then they wonder why the result looks generic, flat, and nothing like what they imagined.

The problem is not Midjourney. The problem is that AI image generators are not search engines. They are creative translators. And like any translator, they can only work with what you give them.

Give them vague input, you get vague output. Give them specific, structured input, you get images that look like they came from a professional photoshoot.

Give them vague input, you get vague output.

This guide teaches you the exact framework that separates generic prompts from great ones.

Why most prompts fail

There are three reasons most Midjourney prompts produce disappointing results:

They describe the subject but nothing else. No lighting, no composition, no camera, no mood.
They use vague adjectives. 'Nice', 'beautiful', 'good lighting' mean nothing to an AI.
They skip the visual language. 'Blue dress' tells AI almost nothing. 'Deep cerulean silk midi dress with cool purple undertones' tells it everything.

The fix is a framework called the 6-layer prompt structure.

The 6-layer prompt framework

Every great Midjourney prompt is built from 6 layers in this exact order:

Each layer adds specificity - by layer 6 the prompt is fully art-directed.

Layer 1 - Content Type

Define what kind of image this is before you describe anything else.

Not just 'photo' - be specific. Portrait. Product still life. Fashion editorial. Flat lay. Cinematic scene. Brand mood visual.

This single decision frames everything that follows. Midjourney interprets 'a woman' completely differently depending on whether the content type is a beauty shot, a street style photo, or a campaign visual.

Example

'Fashion editorial of...' vs 'Beauty studio close-up of...' vs 'Lifestyle portrait of...'

Layer 2 - Subject

Now describe your main subject in full detail. Think like a film director giving instructions to a casting agent and wardrobe department simultaneously.

For people: describe clothing with specific materials and colors, pose energy, hair length and texture, facial expression, and any accessories. Every detail you skip is a detail Midjourney invents randomly.

For products: describe size relative to surroundings, surface texture, finish (glossy, matte, satin), color with specific descriptors, and exact positioning.

Without this

'a woman in a nice outfit'

↓

With this

'a woman in a tailored ivory linen blazer fitted at the waist, relaxed confident stance, loose shoulder-length auburn waves, slight smile, minimal gold hoop earrings'

Layer 3 - Colors and Materials

This is where most prompts fall completely flat.

Hex codes mean nothing to Midjourney. Color names like 'blue' or 'red' are too broad to be useful. What works is descriptive color language that includes temperature, undertone, and finish.

Instead of 'blue' try: deep cerulean with cool gray undertones, powder blue with soft matte finish, midnight navy that absorbs light evenly.

Instead of 'gold' try: warm champagne with subtle satin sheen, antique brass with matte oxidized surface, burnished gold with directional brushed texture.

Apply the same specificity to materials. Silk catches light differently than cotton. Leather has grain and subtle sheen. Velvet absorbs light and creates depth. Name the material and describe how it behaves.

Layer 4 - Composition

Tell Midjourney where to put things and how to frame the shot.

Subject position: centered, rule of thirds, left third, off-center right.
Framing: full body, three-quarter, half body, head and shoulders, close-up.
Background: seamless minimal, contextual environment, soft bokeh, colored backdrop.
Negative space: generous breathing room, tight and intimate, asymmetrical balance.

Without composition direction Midjourney defaults to centered subject with medium framing every single time. This is why so many AI images look the same.

Without composition direction Midjourney defaults to centered subject with medium framing every single time.

Layer 5 - Lighting

Lighting is mood. It is the single most powerful layer in the framework and the most commonly skipped.

Describe three things: the light source, its position, and its quality.

Source: natural window light, studio softbox, golden hour sun, overcast sky, practical lamp.
Position: front-lit, side-lit from the left, back-lit, three-quarter at 45 degrees.
Quality: soft and diffused, hard and directional, dramatic with deep shadows, bright and airy with minimal shadow.

Full example

'soft diffused natural window light from the left, warm golden undertones, gentle shadow falling under the cheekbone, single catchlight in the eye at 12 o'clock position'

That one sentence transforms an image.

Layer 6 - Camera and Lens

The final layer locks in the photographic quality and feel of the image.

Specify a lens. An 85mm portrait lens at f/2.0 creates flattering facial proportions with smooth background bokeh. A 100mm macro lens reveals material texture in extreme detail. A 24mm wide angle creates expansive environmental context.

Specify a camera look. 'Shot on Canon EOS R5' tells Midjourney to render with clinical digital sharpness. 'Shot on 35mm film' tells it to add organic grain, warmer color grading, and natural imperfections.

These details are what make AI images look like they were taken by a professional photographer rather than generated by software.

Putting it all together

Here is the same basic idea written without the framework and with it:

Without and with the framework

Without the framework

'a skincare product on a surface with nice lighting'

↓

With the 6-layer framework

'Product still life of a dark amber glass serum bottle with gold dropper, centered on a white marble surface with warm gray veining, soft diffused natural window light from the left with warm golden undertones, shallow depth of field, shot on 100mm macro lens at f/4.0, clean minimal background with generous negative space --ar 4:5 --v 7 --stylize 200'

The second prompt will produce a result that looks like it belongs in a premium skincare campaign. The first will produce something forgettable.

The quick test

Before you run any prompt, read it aloud. Ask yourself: could someone with no visual context build this scene in their head from your words alone?

If the answer is no - add more detail. Specificity is not optional in AI prompting. It is the entire job.

Specificity is not optional in AI prompting. It is the entire job.

The Midjourney parameters that matter

Once your prompt text is strong, these parameters refine the output:

--ar sets the aspect ratio. Use 4:5 for Instagram, 1:1 for square, 16:9 for landscape, 2:3 for Pinterest.
--v 7 uses the latest Midjourney model (default as of 2025).
--stylize controls how much creative interpretation Midjourney applies. Lower values (50-100) follow your prompt literally. Higher values (400-800) produce more artistic results that drift from the prompt.
--raw disables Midjourney's aesthetic filter and follows your prompt more literally. Use this when you want photorealistic results without artistic enhancement.

Common mistakes to avoid

Saying 'realistic' instead of specifying a camera and lens. The word realistic means nothing. A Canon lens specification means everything.
Using 'beautiful' or 'stunning' as descriptors. These are opinions, not visual instructions. Replace them with specific visual details.
Skipping lighting entirely. Midjourney defaults to flat even lighting when you give it no direction. This is why AI images often look shadowless and artificial.
Writing the prompt as a sentence. AI prompts work better as structured descriptive phrases, not grammatically correct sentences.

Already have a prompt? Optimize it for your specific platform in one click.

The Platform Optimizer rewrites any prompt for Instagram, Pinterest, TikTok, Amazon, and more - correct aspect ratio, composition adjustments, and platform-native visual language, with your subject and brand descriptors kept intact.

Optimize my prompt

Try it now

The fastest way to apply this framework is to use a tool that builds the 6 layers for you automatically.

Our free Midjourney Prompt Builder lets you select each layer from dropdowns and watch the prompt assemble in real time. When you are done, hit Enhance with AI and get 3 variations - subtle, standard, and cinematic - all built on the 6-layer framework.

Or if you already have a prompt that is not working, paste it into our Prompt Fixer. It diagnoses which layers are missing and rewrites the prompt completely.

Try the Prompt Fixer →