Best Prompts for Image Generation with Midjourney

Midjourney is known for producing striking, artistic images. The quality of your results depends heavily on your prompts. This guide explains how Midjourney prompts are structured, gives you copy-and-paste prompt templates for common needs, and shows how to refine images until they look the way you want.

Getting Started with Midjourney

1. Go to the Midjourney website and sign in. Midjourney started on Discord and now also offers a full web interface where you can create and organize images.
2. Choose a subscription. Midjourney has generally required a paid plan, so check the official site for current plans.
3. Open the create page and type a prompt in the prompt bar.

[Screenshot: Midjourney web interface prompt bar]

The Anatomy of a Great Prompt

A reliable structure is:

Subject + details + environment + lighting + composition/camera + style/medium + parameters

• Subject: What is the main focus? Be specific: “an elderly fisherman,” not “a man.”
• Details: clothing, expression, materials, colors.
• Environment: where the scene takes place.
• Lighting: golden hour, soft studio light, neon, overcast.
• Composition: close-up, wide shot, top-down, rule of thirds.
• Style: photograph, watercolor, 3D render, flat vector illustration.

Useful parameters

Parameters are added at the end of the prompt. Long-standing ones include:

• --ar 16:9 sets the aspect ratio (use 9:16 for vertical, 1:1 for square).
• --no excludes elements, for example --no text, watermark.
• --stylize (or --s) controls how strongly Midjourney applies its own aesthetic.
• --chaos increases variety between the generated options.
• --seed helps reproduce similar results.

Parameters and their ranges can change between model versions, so check Midjourney’s official documentation for the version you use.

Best Prompt Templates by Use Case

1. Realistic portrait photography

Portrait of an elderly fisherman with a weathered face and a
knitted wool cap, standing on a foggy harbor at dawn, soft diffused
light, shallow depth of field, 85mm lens, documentary photography
--ar 4:5

2. Product photography

Minimalist product photo of a matte black ceramic coffee mug on a
light oak table, soft window light from the left, subtle shadows,
clean beige background, commercial studio photography --ar 1:1
--no text, logo

3. Blog header illustration

Flat vector illustration of a person working on a laptop
surrounded by floating icons of documents, charts and chat bubbles,
pastel blue and coral palette, clean lines, plenty of negative space
--ar 16:9 --no text

4. Fantasy landscape

Vast floating islands connected by rope bridges above a sea of
clouds, waterfalls spilling into the sky, sunset with warm orange
and violet tones, epic wide shot, highly detailed digital painting
--ar 21:9

5. Logo and icon concepts

Simple geometric logo concept of a fox head made of triangles,
flat design, single color orange on white background, minimal
--ar 1:1 --no gradient, shadow

Note: Treat AI logos as concepts. Have the final logo refined by a designer and check trademark availability.

6. Food photography

Overhead shot of a rustic bowl of ramen with soft-boiled egg,
scallions and steam rising, dark slate table, moody side lighting,
editorial food photography --ar 4:5

7. Isometric 3D scene

Isometric 3D render of a cozy tiny home office with plants,
bookshelves and a glowing monitor, soft pastel colors, clay
material style, studio lighting --ar 1:1

[Screenshot: Four image options from the product photography prompt]

How to Refine Your Images Step by Step

1. Generate a first batch with a simple, clear prompt.
2. Pick the closest result and create variations of it.
3. Adjust one thing at a time: change the lighting, then the composition, then the style. Changing everything at once makes it hard to learn what worked.
4. Use upscaling on your final choice for higher resolution.
5. Use editing tools (such as region editing or extending the canvas, where available) to fix details.
6. Use image references for style or character consistency when your version supports them.

Prompt Mistakes to Avoid

• Too vague: “a nice landscape” gives you random results.
• Contradictions: “minimalist, highly detailed, busy” confuses the model.
• Keyword soup: stacking dozens of words like “8k, ultra, masterpiece” rarely helps with modern models. Clear descriptions work better.
• Expecting perfect text: text rendering has improved, but double-check spelling, or add text later in a design tool.
• Copying living artists’ names to imitate their style raises ethical concerns and may raise legal ones. Describe the style instead.

Using Your Images Responsibly

• Read Midjourney’s current terms on commercial use and visibility. Some plans make your images publicly visible by default.
• Don’t create misleading images of real people or events.
• Label AI images where platforms or clients require it.

Want to bring your images to life? Read How to Create Videos with AI. To write better prompts in general, our ChatGPT writing guide covers giving context clearly.

Worked Example: From a Weak Prompt to a Finished Blog Header

Let’s improve one prompt in four rounds and describe what changes each time.

Round 1 – weak prompt:

AI productivity

Typical result: abstract glowing brains and circuit patterns. Visually busy, generic, and hard to use as a header.

Round 2 – add a subject and medium:

Flat illustration of a person at a tidy desk using a laptop,
calm workspace --ar 16:9

Typical result: much closer, but colors and style may vary a lot between the options.

Round 3 – add palette, composition and exclusions:

Flat vector illustration of a person at a tidy desk using a laptop,
a coffee cup and a small plant, soft pastel blue and coral palette,
subject on the right third, empty space on the left for a headline
--ar 16:9 --no text, logo

Typical result: consistent, clean images with room for your title text, which is exactly what a blog header needs.

Round 4 – refine: pick the best option, create subtle variations, then upscale. If the hands look odd, use the editor to regenerate only that region.

Prompt Building Blocks You Can Mix and Match

Element	Example words
Lighting	golden hour, overcast soft light, studio softbox, neon glow, candlelight
Camera / composition	close-up, wide establishing shot, top-down flat lay, low angle, rule of thirds
Lens feel	35mm documentary, 85mm portrait, macro, wide-angle
Medium	photograph, watercolor, oil painting, flat vector, clay 3D render, pencil sketch
Mood	calm, cozy, dramatic, playful, mysterious, minimalist
Color	muted earth tones, pastel palette, monochrome, high-contrast complementary colors

Pick one or two words from each row rather than stacking many. Fewer, clearer words usually give more predictable results.

Troubleshooting Common Midjourney Problems

1. Images ignore part of the prompt: Put the most important element first and remove less important details.
2. Unwanted text or watermarks appear: Add `–no text, watermark, letters`.
3. Faces or hands look distorted: Choose a less complex pose, use a closer crop, or fix the area with region editing.
4. Style keeps changing between images: Reuse the same style words and consider a style reference or a fixed seed.
5. Colors are too saturated: Add “muted tones” or “natural colors” and lower the stylize value.
6. Too similar options: Increase `–chaos` slightly to get more variety.

Organizing Your Best Prompts

Keep a simple prompt library (a spreadsheet works well) with columns for the prompt, parameters, use case, and a thumbnail of the result. Over time this becomes your own style guide, and it makes it much easier to create consistent images for a website or brand.

Use case: blog header | Prompt: Flat vector illustration of ... |
Params: --ar 16:9 --no text | Notes: left side empty for title

Midjourney Prompt Checklist

• Is the main subject clear and specific?
• Have I described lighting and composition?
• Is the style or medium stated?
• Does the aspect ratio fit where I’ll use the image?
• Have I excluded things I don’t want?
• Am I avoiding real people’s likenesses and living artists’ names?

Ethical and Practical Use on a Website

If you use Midjourney images on a blog or business site, add descriptive alt text for accessibility and SEO, compress images before uploading so pages load quickly, and keep a record of which images are AI-generated in case a client, platform, or advertiser asks.

Bonus Templates for Social Media and Marketing

Bright lifestyle photo of friends sharing a picnic in a city park,
candid laughter, late afternoon sun, natural colors, editorial style
--ar 4:5 --no text

Clean mockup-style scene of a smartphone on a marble desk next to
a notebook and sunglasses, top-down flat lay, soft daylight,
empty screen --ar 1:1 --no logo, text

Seamless repeating pattern of tiny hand-drawn lemons and leaves,
pastel yellow and sage green on cream background, textile design
--ar 1:1 --tile

Pattern and tiling options depend on the model version, so check the documentation if a parameter isn’t recognized.

How to Describe a Style Without Naming an Artist

Instead of naming a specific artist, describe the visual qualities you like: the medium (gouache, ink, oil), the brushwork (loose, precise), the palette (muted, vivid), the era or movement (mid-century poster, Art Nouveau), and the mood. For example, “loose watercolor washes, ink outlines, muted earth tones, vintage travel poster feel” gives Midjourney plenty to work with while keeping your images original.

Frequently Asked Questions

Is Midjourney free?

Midjourney has generally required a paid subscription, with occasional trials. Check the official website for current options.

Can I use Midjourney images commercially?

Paid plans have generally allowed commercial use, with conditions for larger companies. Always read the current terms of service.

Do I still need Discord?

No. Midjourney’s web interface lets you create images directly in your browser, though Discord remains an option.

How long should a Midjourney prompt be?

One or two clear sentences plus parameters is usually enough. Add detail only where it changes the result.