How Midjourney Works: A Practical Guide to Prompts and Parameters

You’ve seen stunning AI-generated images online. 

Chances are, many were made with Midjourney—one of the most popular AI image generators today. 

But how does it work?
 And what do all those weird flags like --v 5 or --ar 16:9 actually do? 

Here’s a simple breakdown.  

What is Midjourney?

 
Midjourney is an AI tool that turns text prompts into images.
You type a short (or detailed) description, and it gives you four visual options.
 
You interact with it via Discord, not a web app.
That’s unique—and also confusing if you’re new.
 
 

How do you use it?

1. Join the official Midjourney Discord
2. Go to any “newbies” channel
3. Type a prompt like:

4. /imagine prompt a cyberpunk city at night, raining --ar 16:9 --v 5 --style raw

5. Wait for your image
6. Click U (upscale) or V (variation) on any of the 4 results

Prompt basics

 
The structure is simple:

Edit/imagine prompt [your description here] [parameters]

Example:

Edit/imagine prompt hyperrealistic photo of a cat wearing sunglasses --v 5 --ar 1:1

Your words guide the content.
 The parameters guide the style, quality, format, etc.

Useful Parameters (Explained)

 
Here are the most common ones:

--v 5

 
Model version.
 Midjourney V5 is the most realistic and flexible.
 You can also try --v 6 (alpha) or older versions if needed.
 

--ar 16:9

 
Aspect ratio.
 Examples:
 

• 1:1 = square
• 16:9 = widescreen
• 9:16 = vertical for mobile

--q 2

 
Quality level.
 

• --q 1 = default
• --q 2 = better detail (takes longer)
• --q .5 = faster, lower quality

--style raw

 
Removes Midjourney’s “artsy” aesthetic.
 Useful when you want photorealism or exact rendering.
 

--no [word]

 
Exclude something.
 Example: --no text removes watermarks or unwanted letters.
 

--seed 12345

 
Makes the output more consistent when rerunning the same prompt.
 

--chaos 0–100

 
Controls randomness.
 Higher = more variation in the 4 images.
 

--stylize 0–1000

 
Controls how strongly Midjourney applies its default style.
 Higher = more stylized, lower = more literal.
 
 

Example prompt breakdown

Edit/imagine prompt an astronaut riding a horse on Mars --ar 3:2 --v 5 --style raw --chaos 40 --no text

This prompt will give you:
 

• A wide image (3:2)
• Realistic style
• Some creative randomness
• No unwanted text overlays

Advanced: Multi-prompts

 
You can also separate concepts using :: to give more weight to certain parts.
 
Example:
 

Edit/imagine prompt cat::2 robot::1 sunset::0.5

Here, “cat” is weighted 2x more than “robot”, and “sunset” is barely considered.
 
 

What Midjourney is not

• It doesn’t understand long dialogue or stories
• It’s not good at spelling words in images
• It doesn’t always follow prompts literally unless you guide it with constraints

Summary

 
To get better Midjourney results:
 

• Learn the core parameters
• Keep prompts short but descriptive
• Use weights, styles, and exclusions to control the output
• Try different versions to see what works for your use case