Midjourney Guide: Create AI Art That Doesn't Look AI

I’ll be upfront: I was a skeptic. When I first tried Midjourney v4 in 2023, I thought “this is a gimmick.” The images were cool but they all had that look — you know the one, where you can tell it’s AI from across the room.

Then v5 dropped. Then v6. And suddenly I was using it to generate featured images for blog posts, mockups for client presentations, and — embarassingly — a custom birthday card for my mom (she loved it, thought I’d hired a designer).

This guide is everything I wish I’d known when I started. It won’t make you a pro in an afternoon, but it’ll save you the two months of confusion I went through.

First: Discord or Web?

Midjourney started on Discord, and if you’re over 30, that sentence probably fills you with dread. Discord is… a lot. Channels moving, messages scrolling, other people’s generations clogging up your view.

Use the web interface. Go to midjourney.com, log in, and use the web generator. It’s cleaner, you can actually see your image history in a gallery view, and it feels like a proper tool instead of a chat room.

That said, Discord is still useful for one thing: seeing what other people prompt. There’s a lot of value in browsing the #showcase channel and stealing — politely, “finding inspiration from” — other people’s prompt structures. Just don’t copy-paste someone’s prompt and call it your own. That’s not how creativity works.

The Pricing Reality (Yes, You Need to Pay)

Midjourney isn’t free. And honestly? That’s fair. Running GPUs costs real money.

Plan	Price	What You Get	Worth It?
Basic	$10/mo	~200 images, public only	Good for testing
Standard	$30/mo	Unlimited relaxed + fast hours	Start here
Pro	$60/mo	+ stealth mode	If you’re doing client work
Mega	$120/mo	Maximum fast hours	Power users only

My take: start with Basic for a week. Generate 20-30 images. If you find yourself wanting more, upgrade to Standard. The “relaxed” generation mode means you can queue up images in the background while you do other work — they take longer but they don’t eat into your fast hours.

One thing that annoyed me: you can’t “pause” your subscription mid-month. So if you sign up for Standard and realize you’re not using it, you’re stuck until the month rolls over. Cancel immediately if you’re not using it.

Writing Your First Prompt (and Why “A Cat” Won’t Cut It)

The biggest mistake beginners make: under-describing. They type “a cat in a park” and wonder why the result looks like generic stock photography.

Midjourney thrives on specificity. It’s not that it needs more words — it’s that it needs interesting words.

What didn’t work for me:

/imagine prompt: a dog

Cute, sure. But also generic. I could’ve found that on Unsplash in ten seconds.

What actually worked:

/imagine prompt: a golden retriever mid-shake, water droplets
frozen in air, golden hour backlighting, shot on Canon 5D
Mark IV, 85mm lens, f/1.4, motion blur on the ears
--ar 3:2 --style raw

That’s specific. That has texture. And Midjourney delivered something that actually looked like photography, not “AI art.”

The Formula I Use

After a lot of trial and error, here’s the prompt structure that consistently works:

1. Subject — what’s the main thing?
2. Modifier — adjective or style that changes the feel
3. Technical spec — camera, lens, lighting (even for digital art, this helps)
4. Parameters — aspect ratio, style raw, etc.

You don’t always need all four. But if your prompt is just “a castle,” add at least one more layer.

Parameters Cheat Sheet (The Ones I Actually Use)

Midjourney has a ton of parameters. Most of them, I never touch. Here are the ones that show up in my prompts every week:

Parameter	What It Does	When I Use It
--ar 16:9	Changes aspect ratio	Blog headers, thumbnails
--ar 9:16	Vertical	Social media stories
--style raw	Less “AI artistic,” more realistic	Photos, product shots
--stylize 250	Controls artistic intensity (0-1000)	Lower for realism, higher for art
--no text, watermark	Excludes elements	Every. Single. Time.
--seed 12345	Reproduces the same “random” look	When I find a style I like
--v 6	Model version	Always use the latest

A few notes on these:

--style raw is underrated. Midjourney’s default style adds artistic flair whether you want it or not. If you’re going for photorealism, --style raw is mandatory. Without it, your “photo of a coffee shop” will look like a painting of a coffee shop.

--stylize matters more than you think. At 0, Midjourney follows your prompt closely but the image might look dull. At 1000, it looks amazing but might ignore half your prompt. I usually land around 200-400 for a good balance.

--no is your friend. I have --no text, watermark, blurry, low quality, extra limbs saved in a Notes file and I paste it into almost every prompt. It’s not that Midjourney always adds these things — it’s that when it does, it ruins the image.

Advanced Stuff That’s Actually Useful (Not Just Show-Off)

Style References (--sref)

This is genuinely powerful. You can take an image you like — maybe a movie still, maybe an oil painting — and tell Midjourney “use the style of this image for my new prompt.”

/imagine prompt: a futuristic Tokyo street market --sref [URL_of_image_you_like] --ar 16:9

The result inherits the lighting, color palette, and artistic style of your reference — but applies it to your new subject. I’ve used this to create a series of blog header images that all “feel” like they belong together. Game-changer for branding.

Character Consistency (--cref)

If you’re creating a character and want them to look the same across multiple images (useful for stories, comics, or consistent avatars), use --cref.

/imagine prompt: a young woman with short purple hair, exploring
a neon-lit alleyway --cref [URL_of_character_image] --ar 16:9

The --cw parameter controls how closely it matches — 100 matches everything (face, hair, clothing), 0 matches just the face. I usually set it around 50-70 so the character is recognizable but the outfit can change per scene.

Multi-Prompting with Weights

You can blend concepts by giving them weights:

/imagine prompt: cyberpunk cityscape::2 rain::1 --ar 16:9

The ::2 means “twice as important as default.” So this image will lean cyberpunk but still have noticeable rain. It’s a subtle way to steer the composition without rewriting your entire prompt.

The Mistakes I Made (So You Don’t Have To)

Mistake #1: Making Prompts Too Long

There’s a temptation to describe everything. “A cat sitting on a windowsill, orange tabby, morning light, with a coffee cup in the foreground, and a plant in the background, and maybe some bokeh…”

Stop. Midjourney gets confused if you try to cram in 15 different elements. Focus on 3-5 strong ones. You can always use variations (V1, V2, etc.) to explore different compositions.

Mistake #2: Not Iterating

Your first generation is rarely your best. Use the V (variation) buttons to explore different versions of the image you like. Then upscale (U1-U4) your favorite.

I probably generate 20-30 variations before I find the one for anything important. That’s normal. Midjourney is a numbers game — the more you generate, the better your odds.

Mistake #3: Ignoring Aspect Ratio

The default is square (1:1). For everything. If you’re making a blog header, that square image is going to look ridiculous. Set --ar 16:9 (or 21:9 for ultra-wide) every time.

Similarly, if you’re generating for Instagram Stories or TikTok, use --ar 9:16. It sounds obvious but I’ve forgotten this so many times and ended up with a square image I had to awkwardly crop.

Mistake #4: Expecting Photorealism Without --style raw

I mentioned this above but it bears repeating. If you want your image to look like a photo, add --style raw. Without it, Midjourney will “artistically enhance” your image and it’ll look… fine, but not real.

What I Actually Use Midjourney For (Real-World Use Cases)

Blog Post Featured Images

This is 80% of my Midjourney use. I need a header image for an article about productivity tools — I’m not going to find a stock photo that says “productivity tools” in a way that’s not boring.

abstract representation of AI and productivity, clean minimal
workspace, soft blue and warm gray palette, subtle geometry,
no text, suitable for blog header --ar 16:9 --stylize 150
--style raw

Clean, professional, and unique. No stock photo licensing to worry about.

Social Media Content

For Instagram or LinkedIn, I’ll generate backgrounds or mood images:

flat lay of a productive desk setup, laptop, coffee, notebook,
top-down view, bright natural light, lifestyle photography
--ar 4:5 --style raw

These work great as post backgrounds with text overlay.

Client Presentations (Concept Visualiation)

If I’m explaining a concept to a client — say, “imagine a dashboard that looks like this” — I can generate a mockup-quality visualization in minutes instead of spending an hour in Figma.

It’s not a replacement for real design. But for communicating an idea, it’s unbeatable.

Midjourney vs. DALL-E 4 vs. Stable Diffusion

People ask me which one to use. Here’s my honest take:

• Midjourney if you care about aesthetics. It produces the most beautiful images out of the box. The artistic sensibility is genuinely impressive.
• DALL-E 4 (via ChatGPT) if you need accuracy. It follows prompts more literally. If you need “a red apple on a blue table,” DALL-E will give you exactly that. Midjourney might give you “a stylized interpretation of an apple.”
• Stable Diffusion if you want control. It’s open-source, you can run it locally, and you can fine-tune it on your own images. But it has a steep learning curve and the results require more tweaking.

For most people reading this: start with Midjourney. It’s the easiest to use and the results are genuinely impressive even with simple prompts.

A Few Final Thoughts

Midjourney is fun. That’s the part that got me hooked — not the “productivity” angle, not the “content strategy” angle. It’s genuinely enjoyable to type in a weird idea and see what comes out.

But it’s also a real tool. I’ve saved hours of searching for stock photos. I’ve created visuals for articles that would’ve cost $200+ to commission from a designer. And yeah, I’ve made some cool art for my apartment walls.

One last thing: respect the artists. Midjourney was trained on real human art, and there’s an ongoing debate about consent and compensation. I don’t have a clean answer for you on that — just… be aware of it. Use Midjourney for ideation, exploration, and personal projects. If you’re using it commercially, think about whether what you’re making adds value or just replaces human creativity entirely.

Now go generate something cool. And when you make something you love, screenshot it — Midjourney’s web gallery is fine, but I’ve lost images before and it suuucks.

Frequently Asked Questions

How much does Midjourney cost?

Midjourney offers four plans: Basic ($10/month), Standard ($30/month), Pro ($60/month), and Mega ($120/month). The Standard plan is recommended for most users.

Do I need Discord to use Midjourney?

Midjourney now has a web interface at midjourney.com, but Discord remains fully supported. Both platforms offer the same functionality.

Can I use Midjourney images commercially?

Yes, paid subscribers can use generated images commercially. The Pro and Mega plans include stealth mode for private generation.