How to Use Midjourney for Beginners: From Zero to Your First Image

The first time I opened Midjourney, I closed the tab within five minutes. It was on Discord. Everything was a slash command. Strangers' weird AI portraits scrolled past constantly. I had no idea where my own images would even appear.

A week later I came back, pushed through the friction, and made my first decent image. Now I use it almost daily for blog covers, social posts, and the occasional weird experiment. The tool is genuinely good — it's just terrible at onboarding new people.

This guide is the version I wish someone had handed me on day one. Skip the YouTube videos with 47 prompt formulas. You only need a few things to get going.

Step 1: Sign up and pay (5 minutes)

Midjourney has no free plan anymore. You'll need at least the Basic plan at $10/month, which gives you about 200 images.

1. Go to midjourney.com and sign in with your Discord or Google account
2. Subscribe to the Basic plan from the Manage Sub page
3. Either use the web app at midjourney.com (recommended for beginners) or join the Midjourney Discord server

The web app launched in 2024 and is much easier than Discord. Use it. The Discord workflow is mostly historical at this point.

If you're already on Discord and want to use it there, you'll need to direct-message the Midjourney bot or use it in a server channel. Don't try to learn in the public newcomer channels — your prompts will scroll off-screen in 10 seconds.

Step 2: Understand what a prompt actually is

A Midjourney prompt has three parts. You only need the first one to start.

Part 1 — The subject (required):

a cozy cabin in a snowy forest at sunset

Part 2 — Style modifiers (optional):

cinematic, soft lighting, photorealistic, 35mm film

Part 3 — Parameters (optional):

--ar 16:9 --v 6.1

Most people copy massive prompts from Twitter and assume that's the only way. It's not. Short, clear prompts work great. The 200-word prompts you see online are usually for very specific stylistic effects, not general image quality.

Step 3: Make your first image (3 minutes)

In the web app, click the Imagine bar at the top and type:

a cozy cabin in a snowy forest at sunset

Hit enter. Wait about 60 seconds. You'll get four image variations.

That's it. You've made your first Midjourney image.

Step 4: Learn the four buttons under every image

Once your four images appear, you'll see buttons under them. The four that actually matter:

• U1, U2, U3, U4 — Upscale variation 1, 2, 3, or 4 (gives you a bigger, cleaner version of one of the four)
• V1, V2, V3, V4 — Generate four new variations based on one of the originals
• 🔄 (Re-roll) — Generate a completely new set with the same prompt
• Vary (Subtle/Strong) — After upscaling, generate slight or significant variations

The workflow most people land on:

1. Generate four images
2. Pick the closest one
3. Click V on that one to get four variations of it
4. Repeat until one is right
5. Upscale (U) the winner

This is faster than rewriting your prompt 12 times trying to nudge the model in a direction.

Step 5: The five things that improve any prompt

These are the small additions that make a beginner's images suddenly look professional.

1. Specify the medium. "Photo of," "oil painting of," "pencil sketch of," "3D render of." This single change has the biggest impact on the visual style.

2. Add a lighting term. "Golden hour," "soft window light," "studio lighting," "blue hour," "neon-lit." Lighting separates great images from flat ones.

3. Include a camera or lens detail (for photos). "Shot on 35mm film," "shallow depth of field," "wide-angle lens," "85mm portrait lens." Even when these don't literally apply, they shift the model toward photographic-looking output.

4. Add a mood or atmosphere word. "Moody," "dreamy," "tense," "serene," "chaotic." This guides the emotional read of the image.

5. Pick an aspect ratio. The default is square (1:1). For most use cases, you want different:

• --ar 16:9 for blog covers and YouTube thumbnails
• --ar 9:16 for vertical social posts (Reels, TikTok, Stories)
• --ar 4:5 for Instagram feed
• --ar 3:2 for general photo aspect

Add the --ar flag at the end of your prompt.

A real before-and-after

Beginner prompt:

a coffee shop

Output: generic stock-photo-looking interior, square format, no atmosphere.

Improved prompt:

photo of a small coffee shop interior, morning light through large windows, warm tones, shallow depth of field, shot on 35mm film, cozy atmosphere --ar 16:9

Output: editorial-quality image you'd actually use in a blog post.

The second prompt isn't longer because it's clever. It's longer because it specifies the medium (photo), the lighting (morning, warm), the technique (shallow depth of field, 35mm), and the aspect ratio. Each addition is functional.

Step 6: Use image references for consistency

This is the feature that turns Midjourney from "generate a random image" into "make this thing fit my brand."

Drag any image into the Imagine bar. It uploads. Then write a prompt like:

[image] in the style of a vintage travel poster, warm color palette --ar 3:2

Midjourney uses the uploaded image as a stylistic reference. You can also use the --cref (character reference) or --sref (style reference) parameters with image URLs for more specific control.

If you're making a series of images for one blog post or campaign, this keeps them visually consistent. Pick one image you love, then reference it in every subsequent generation.

Step 7: When things go wrong

The four most common beginner problems and what to do about them:

Problem: Faces look distorted or weird. Add "professional portrait" or "studio photography" to the prompt. Avoid generating images with multiple people unless you specifically need a group — Midjourney handles single subjects much better.

Problem: The text in the image is gibberish. This is a known limitation. Midjourney is steadily improving at text but is not reliable for it. For images that need readable text (like ads, posters with quotes, etc.), generate the image without text and add the text in Canva or Figma afterward.

Problem: It keeps giving me the same style I don't want. Try the opposite. If everything looks too photorealistic, add "illustration" or "watercolor." If everything looks too stylized, add "photo" and a camera detail. Sometimes you need to override a default style rather than just refining it.

Problem: Images aren't showing what I described. Simplify. If you're asking for "a knight on a horse riding through a fantasy forest with three dragons in the sky during sunset" — pick three of those elements, not all six. Midjourney handles 2–4 main subjects well. More than that and things blur together.

What I use Midjourney for now

After about 6 months of regular use, here's what I actually use it for:

• Blog post cover images (probably 80% of my use)
• Social media post graphics
• Mood boards for design projects
• Quick illustrations for client decks
• The occasional weird "what if" experiment

What I don't use it for:

• Anything that needs a real, specific human face (use stock photos or actual photography)
• Anything that needs accurate text in the image
• Logos or icons (use a vector tool)
• Diagrams or infographics (it's bad at structured information)

It's a great tool when you know what it's good at. Like every AI tool, it gets worse the further you push it from its sweet spot.

The takeaway

Don't start by reading prompt formulas on Twitter. Start by typing five-word prompts and learning what comes out.

After 30 minutes of just generating and varying, you'll understand the model's defaults better than any guide can teach you. From there, the prompt techniques in this post will make sense — because you'll have already seen what they're correcting for.