How to Use Midjourney: Complete Beginner's Guide

What Is Midjourney — and Why Is It Different?

Midjourney is an AI image generation model accessible through Discord. Unlike DALL-E (web-based) or Stable Diffusion (local install), Midjourney runs entirely in a Discord server, which is unusual but has advantages: a thriving community, instant access to others' prompts for inspiration, and no local GPU required.

Midjourney consistently produces the most aesthetically polished images of any AI image generator — particularly strong at photorealism, painterly illustration, and cinematic compositing. It tends to be preferred by professionals for client work, marketing assets, and concept art.

Getting Started on Discord

1. Create a Discord account at discord.com if you don't have one
2. Go to midjourney.com and click "Join the Beta" — this invites you to the Midjourney Discord server
3. Subscribe to a plan at midjourney.com/account. The Basic plan ($10/month) gives ~200 image generations. The Standard plan ($30/month) gives unlimited "relaxed" generations
4. Once subscribed, you can use Midjourney in any Discord server where the bot is installed, or in direct messages with the bot
5. To DM the bot: find Midjourney Bot in the member list and send it a message

The /imagine Command: Basic Syntax

Every image starts with /imagine followed by a text prompt:

/imagine a red fox sitting in a snowy forest, golden hour light, photorealistic

Midjourney generates 4 image variations by default. Below the grid you'll see buttons:

• U1–U4: Upscale a specific image (higher resolution, more detail)
• V1–V4: Create 4 variations of a specific image (similar but not identical)
• 🔄: Regenerate — run the same prompt again for new variations

Aspect Ratios: --ar

By default Midjourney generates square (1:1) images. Change this with --ar:

• --ar 16:9 — landscape, great for YouTube thumbnails, desktop wallpapers, hero images
• --ar 9:16 — portrait, perfect for Instagram Stories, TikTok, phone wallpapers
• --ar 4:3 — classic photo ratio
• --ar 2:3 — portrait photography, book covers
• --ar 1:1 — square, best for social media posts, logos

/imagine aerial view of Tokyo at night, neon lights, rain --ar 16:9

Stylize Parameter: --stylize (or --s)

Controls how much Midjourney's aesthetic training influences the image. Range: 0–1000. Default: 100.

• --stylize 0: Follows your prompt very literally, minimal AI "artistic interpretation"
• --stylize 100: Default balance of prompt accuracy and aesthetic quality
• --stylize 750: More artistic, more opinionated, may deviate from prompt
• --stylize 1000: Maximum artistic freedom — beautiful but may ignore parts of your prompt

Style Raw: --style raw

Adding --style raw reduces Midjourney's tendency to add its signature polished look. This is useful when you want:

• More photorealistic images without the "Midjourney look"
• Images that match real-world photography more closely
• Less over-saturated, less stylized outputs

/imagine portrait of a woman, natural window light, Canon 5D, 85mm lens --style raw --ar 2:3

Version Parameter: --v

Always use the latest version for best results. As of 2025, that's --v 6.1. You can set a default version in /settings so you don't have to add it every time.

Quality Parameter: --q

--q 2 doubles the rendering time but produces more detailed images. --q .5 is faster and cheaper. For final outputs, use --q 2. For quick exploration, use the default or --q .5.

Negative Prompting: --no

Tell Midjourney what to exclude from the image:

/imagine a mountain landscape, dramatic sky --no people, cars, buildings, text

Common uses: --no text, watermarks (remove unwanted text), --no blurry, out of focus (force sharpness), --no cartoon, illustration (force photorealism).

5 Example Prompts with Expected Results

1. Cinematic Photorealism

/imagine close-up portrait of an elderly fisherman, weathered face, deep-set eyes, fog in background, shot on Leica M11, 50mm Summilux, golden hour --ar 2:3 --style raw --stylize 200

Produces a hyper-realistic portrait with filmic quality, deep texture in skin, and moody atmosphere.

2. Fantasy Concept Art

/imagine ancient library built inside a giant tree, glowing bookshelves spiraling upward, shafts of magical light, detailed illustration, Artstation trending --ar 16:9 --stylize 750

Produces richly detailed fantasy concept art with warm lighting and intricate architectural detail.

3. Product Photography

/imagine luxury minimalist perfume bottle, dark background, single dramatic light source, reflections on surface, commercial product photography --ar 1:1 --style raw --q 2

Clean, professional product shot suitable for use in marketing materials.

4. Flat Illustration

/imagine isometric city illustration, pastel colors, tiny detailed buildings and streets, flat design style, clean background --ar 1:1 --stylize 300 --no shadows, gradients

Produces a vector-style isometric illustration with consistent colors — great for app icons, blog headers.

5. Abstract Art

/imagine fluid ink in water, deep blue and gold, macro photography, swirling patterns, high contrast, abstract --ar 16:9 --stylize 900 --q 2

Produces mesmerizing abstract imagery with fine detail throughout.

Prompt Structure Formula

This formula reliably produces good results:

[Subject], [Setting/Background], [Mood/Lighting], [Style/Medium], [Camera/Technical], [Parameters]

Example applying the formula:

a corgi wearing a tiny hat [subject], sitting on a wooden stool in a sunlit kitchen [setting], warm afternoon light, cheerful [mood/lighting], watercolor illustration [style], --ar 4:3 --stylize 400

Common Beginner Mistakes

• Too many concepts: "a dragon flying over a futuristic city at sunset with cherry blossoms and a samurai fighting robots" — Midjourney will try to include everything and the result is chaotic. Focus on 2–3 key elements.
• No style direction: Saying "a dog" gives you a random style. Add "photorealistic", "oil painting", "anime illustration", or "3D render" to control the aesthetic.
• Ignoring --ar: Most compositions look much better with appropriate aspect ratios. Portraits need tall ratios; landscapes need wide ones.
• Not upscaling: The initial 2x2 grid is low resolution. Always upscale (U button) the image you want to use.
• Over-relying on "realistic": Use specific technical language instead — "shot on Canon EOS R5, 85mm f/1.4, shallow depth of field" is far more effective than just "realistic".