Midjourney prompts work best when they read like intentional visual direction. A prompt should establish the subject, scene, composition, and treatment before it accumulates decorative adjectives. Parameters can influence format and variation, but they do not rescue an unclear idea.
Features and parameter behavior can change, so verify current syntax in Midjourney’s own documentation before building a production workflow around a specific option. The principles in this guide are deliberately durable: describe what matters, control one variable at a time, and judge results at their intended size.
A stable prompt structure
Build the descriptive part in this order:
Main subject and action, setting, composition and viewpoint, lighting, medium and style characteristics, color and texture, critical constraints
Example:
Compact reading chair with a curved walnut frame and oatmeal upholstery, placed beside a tall window in a quiet apartment corner, medium eye-level product view with the chair on the right third and open wall on the left, soft overcast daylight, refined interior editorial photography, warm neutral palette, realistic wood grain and woven fabric, uncluttered scene, no people, no text
The subject appears first because it owns the image. Placement and open wall are explicit because the layout matters. Material accuracy is more actionable than “luxurious.”
Decide what the prompt should control
Before writing, divide your requirements into three groups:
- Essential: removing this would change the purpose of the image.
- Preferred: this improves the direction but can vary.
- Open: the model can interpret this freely.
For a book-cover concept, portrait orientation and open title space may be essential. Clothing details might be preferred. Background plants may remain open. This hierarchy keeps the prompt from becoming a flat list where every word competes equally.
Composition language that helps
Use observable framing terms:
- close-up, medium view, full-length, wide establishing view;
- eye level, low angle, overhead, top-down, three-quarter view;
- centered symmetry, left-third placement, diagonal movement;
- shallow depth of field, layered foreground, flat graphic space;
- negative space above, simple area on the right, edge-to-edge pattern.
Lens language can suggest a visual effect, but it should not become false precision. “Compressed portrait perspective” or “wide environmental view with straight verticals” may communicate more than a random focal length.
Style without imitation
Describe the characteristics you want instead of depending on an artist’s name. For an illustration, specify line weight, shape language, texture, palette, and rendering depth. For photography, specify light behavior, contrast, grain, material realism, and editorial context.
Compare:
whimsical art
with:
hand-painted gouache illustration, simplified rounded shapes, visible dry-brush texture, muted ochre and forest-green palette, gentle asymmetry, cream paper background
The second prompt provides visible decisions that can be evaluated and revised.
Parameters: use them after the concept
Midjourney supports parameters for aspects such as image ratio and variation. Syntax and availability can change between versions. Add only parameters that serve the final use, and place them where the current interface expects them.
An aspect ratio is useful because format strongly affects composition. A landscape website header, vertical poster, and square product tile should not share the same layout. Model version or stylization controls can matter too, but compare them using the same core prompt so you understand what the control changed.
Do not paste a long inherited parameter string into every job. Unsupported or contradictory controls create confusion and make the prompt harder to maintain.
A disciplined iteration sequence
- Test the core idea. Use the subject, setting, and broad treatment.
- Correct composition. Adjust crop, placement, viewpoint, or negative space.
- Correct material and light. Add specific surfaces, light direction, and contrast.
- Remove recurring distractions. Add a short constraint only for problems you actually observe.
- Evaluate at final format. A thumbnail must work small; a poster needs room for type.
Changing everything between runs makes it difficult to learn. A one-variable comparison gives you evidence about which phrase influenced the result.
Prompt templates
Environmental portrait
Environmental portrait of [subject] performing [action] in [setting], [camera distance and viewpoint], subject placed [position], [foreground/background relationship], [light direction and quality], [photographic treatment], [palette], realistic [important textures], [expression or gesture], no readable text, no unrelated props
Graphic poster concept
Vertical poster illustration about [idea], one dominant [symbol or subject], bold [composition], limited [palette] color system, [shape and texture language], clear open area at [location] for later typography, high contrast at small size, no generated words, no logos
Product image
[Product] on [surface/background], [view and placement], [lighting setup], accurate [materials and construction], [commercial mood], realistic contact shadow, simple supporting prop [if needed], no additional products, no invented branding, no text
To turn an existing visual reference into this structure, start with the Image-to-Prompt Generator, choose the Midjourney mode, and remove any detail that your new concept does not need.
If your goal is ecommerce or commercial imagery, continue with the Midjourney product photography prompts guide for catalog shots, lifestyle scenes, detail images, and campaign layouts.
Common mistakes to avoid
Putting the main subject too late. Lead with the image’s owner, not a page of atmosphere.
Repeating synonyms. “Moody, atmospheric, dramatic, cinematic” consumes attention without clarifying the actual light.
Using conflicting media. If you combine photography, watercolor, 3D rendering, and flat vector art, explain the intended hybrid or choose one.
Relying on a parameter instead of composition. Format controls the canvas, not the quality of your visual hierarchy.
Ignoring artifacts around text and hands. Inspect critical anatomy, lettering, labels, and product details before publishing.
How to write a Midjourney prompt
A useful Midjourney prompt starts with the visual idea, not a pile of style words. Write the subject first, then add the setting, composition, light, material or medium, and only the constraints that matter.
A simple structure is:
[subject and action] + [setting] + [composition] + [lighting] + [visual treatment] + [important constraints]
For example:
A lone cyclist crossing a quiet city bridge just before sunrise, wide environmental composition, soft blue-hour light with a warm horizon, realistic editorial photography, subtle road reflections, natural proportions, no visible logos or text.
This gives the model a clear scene to build. If the first result is weak, change the biggest problem first instead of rewriting every part of the prompt.
Using an image as a Midjourney prompt reference
When your starting point is a picture, first separate what you want to preserve from what you want to change. Describe the reference in terms of composition, palette, light, camera distance, materials, and mood rather than trying to copy every pixel.
You can use the Image-to-Prompt Generator to turn a reference into a structured description, then rewrite that description around your own subject and goal.
Keep version-specific syntax and controls separate from the core creative brief. Interfaces and parameters can change, but a precise description of the intended image remains useful.
Frequently Asked Questions
Do longer Midjourney prompts create better images?
Not automatically. Extra words can add control or create competition. Add detail when it changes a decision you care about.
Where should parameters go?
Follow the current Midjourney documentation for supported syntax and placement. Keep the natural-language creative brief readable before adding controls.
Can I use a reference image with a text prompt?
Midjourney workflows can support image references, but the exact controls can evolve. Explain whether the image is guiding content, style, character, or composition, and verify current documentation.
What is the fastest way to improve a weak result?
Identify the largest failure—usually subject clarity, composition, or light—and revise that one dimension. Avoid adding a generic wall of adjectives.