How to Use Midjourney: 10 Steps, 75 Min [2026]

Midjourney turns a written description into a finished image in under a minute. It’s the tool most people still mean when they say “AI image generator,” commanding roughly 26.8 percent of user preference among AI image tools in 2026, according to Morphed.app, and the V8.2 model is its default engine today after replacing V8.1 on July 24, 2026, per VidMuse.ai and a Blake Crosley guide tracking the platform’s version history, with access to the new model included at no extra fee across every plan from $10 to $120 a month. V8.2 became available to everyone on August 27, 2026, adding instruction-based editing, inpainting, and outpainting, reported AlphaSignal, and by September 2026 its editing tools had grown to support plain-language instructions and up to four reference images alongside personalization, moodboards, and style references, according to Midjourney and Releasebot. Reaching it no longer requires Discord. Midjourney now runs a full web app at midjourney.com, with Discord available as an optional sign-in method instead of the only door in.

This tutorial walks through the entire process in order: creating an account, picking a subscription tier that fits your budget, writing prompts that actually produce what you pictured, and using the parameters that turn a lucky roll into a repeatable workflow. Along the way you’ll build a small real project, a set of six branded header images with a locked-in style, so you finish with a working file set instead of a single lucky screenshot.

Set aside about 75 minutes for a full first pass through the steps below, and budget at least $10 a month once you subscribe. Midjourney has no ongoing free tier for new web or Discord accounts. Its only free trial lives inside a separate app called niji・journey. That means you’ll need a card on file before your first prompt, but you won’t need coding experience or a local GPU. Every image renders on Midjourney’s own servers.

·

Google · Preferred Sources

Don't miss new tech stories on Google

Add Tech Insider once in the Google app and our stories appear in your news suggestions.

Add Now

Prerequisites: What You Need Before You Start

Gather these before you open midjourney.com. Nothing here is optional if you want to follow every step in this tutorial without a mid-lesson detour.

• A working email address, or a Google or Discord account you’re willing to use for sign-in
• A credit or debit card, since new accounts require a paid plan starting at $10 a month
• A modern browser (Chrome, Firefox, Safari, or Edge), current within the last two major versions
• Optional: the Discord desktop or mobile app, only if you prefer generating images inside a Discord server rather than the web app
• 75 to 90 minutes to work through every step, plus extra time later for the six-image project near the end
• No design background required. The six-part prompt formula in Step 4 does most of the heavy lifting

If you’d rather avoid a subscription entirely, two other tutorials on this site cover free or lower-cost paths: running FLUX locally in ComfyUI on your own GPU, and Google’s Nano Banana Pro, which bills per image instead of per month. Midjourney still wins on prompt control and community tooling, which is why it remains the default recommendation for most people learning AI image generation for the first time.

Step 1: Create a Midjourney Account and Choose a Plan

Go to midjourney.com and click Sign Up. You can register with an email address, a Google account, or by connecting Discord, and all three lead to the same web app once you’re in. There’s no separate “developer” or “creator” tier to hunt for. Every new account lands on the same subscription menu, and you can change plans at any time from your account settings without losing past generations.

Midjourney pricing hasn’t changed dramatically since paid subscriptions first launched at $10 a month back in 2025, according to BloggersIdeas, and subscriptions still started at that same $10-a-month entry point as of September 2026, AlphaSignal confirmed, even as the platform’s registered user base grew to roughly 20 million accounts, of which about 1.4 million are paying subscribers as of an August 11, 2026 QAI estimate. Four tiers still cover nearly every use case, from casual weekend projects to agency-scale output, and VideoProc’s pricing rundown, updated September 16, 2026, confirms the same lineup remains current: Basic at $10, Standard at $30, Pro at $60, and Mega at $120 a month, all now with V8.2 access included. Here’s what each one includes as of late 2026.

Plan	Monthly Price	Annual Price	Fast GPU Hours	Best For
Basic	$10/mo	$96/yr (~$8/mo)	3.3 hrs	Casual use, light experimentation
Standard	$30/mo	$288/yr (~$24/mo)	15 hrs	Regular creators, unlimited Relax Mode
Pro	$60/mo	$576/yr (~$48/mo)	30 hrs	Stealth Mode, higher-volume commercial work
Mega	$120/mo	$1,152/yr (~$96/mo)	60 hrs	Teams and agencies with heavy output

Annual billing knocks roughly 20 percent off every tier, a discount Ring/lyo130 still confirmed across all four plans as of August 12, 2026, which matters if you already know you’ll stick around past month one. “Fast GPU hours” are the minutes Midjourney spends actually rendering your prompts at full speed, and Basic’s $10 a month buys roughly 200 fast GPU minutes on v7 and newer models, according to a July 4, 2026 SkyCrumbs breakdown. Once you use up your monthly allotment on Standard, Pro, or Mega, generations don’t stop, they just switch to Relax Mode, which queues your jobs at no extra cost but can take several minutes during busy periods, or you can buy extra Fast GPU time outright for about $4 an hour on any tier, a rate Wink.ai’s August 19, 2026 pricing page and AlphaSignal both still had pegged at $4/hour in September 2026, with purchased time carrying no expiration date, unlike your monthly plan allotment. Basic subscribers don’t get Relax Mode, so once those 3.3 fast hours are gone for the month, you’re either paying for extra time, waiting for the reset, or upgrading.

For this tutorial, Standard is the practical starting point. It’s the cheapest plan with Relax Mode, and, per that same July 4, 2026 SkyCrumbs rundown, it’s also the entry point for Draft Mode, the faster rendering option Midjourney first introduced back in April 2025 alongside the V7 launch, on v7 and later models, so at $30 a month and up you can experiment freely with the six-part prompt formula in Step 4 without watching a countdown timer. You can always downgrade to Basic or upgrade to Pro once you know how much you actually generate in a typical month.

What About niji・journey?

If you scroll through Midjourney’s sign-up options, you’ll also spot niji・journey, a separate anime and illustration-focused app built by Midjourney in partnership with Spellbrush, now running on the Niji 7 model the companies rolled out in January 2026 with sharper prompt adherence and notably cleaner detail, including clearer eyes, reflections, and background elements than the version it replaced, according to Midjourney’s own release notes. It’s worth knowing about because it’s currently the only place Midjourney offers a free trial. The core model underneath is tuned specifically for anime, manga, and stylized character art rather than the photorealistic and general-purpose output this tutorial focuses on. If your project is illustration-first, it’s worth testing niji’s free trial before committing to a paid Midjourney plan. For photography-style work, product shots, or anything closer to real-world imagery, the main Midjourney model covered in the rest of this guide is the better fit.

Step 2: Access Midjourney on the Web or in Discord

Once your plan is active, you have two ways to generate images, and they produce identical results. The web app at midjourney.com is the faster on-ramp for most beginners. Log in, click the prompt bar at the top of the Create page, type a description, and press enter. There’s no bot to learn and no server to join.

The Discord route still works, and some longtime users prefer it because it keeps a running, searchable history of every prompt in one channel. Join the official Midjourney Discord server, find one of the “newbies” channels, and type a slash command directly into the message box:

/imagine prompt: [your description here]

Discord autocompletes the slash command as you type, so you don’t need to memorize the exact syntax. Either path draws from the same subscription and the same fast hours, so pick whichever interface fits how you already work. If you’re generating images as part of a team, the web app’s shared organizing tools make collaboration considerably easier than scrolling a shared Discord channel.

Step 3: Generate Your First Image with /imagine

Time to generate something. Keep your first prompt simple so you can see how Midjourney interprets plain language before you start layering on parameters. In the web app’s Create bar, or in Discord after /imagine prompt:, type this:

/imagine prompt: a red fox sitting in a snowy pine forest at golden hour, cinematic lighting, shot on 35mm film, shallow depth of field --ar 16:9 --stylize 250 --v 7

Midjourney’s backend renders images at a rate of roughly 20 to 40 per second across more than 10,000 servers, according to a 2025 infrastructure breakdown from BloggersIdeas, and on the current V8.2 model that throughput translates to renders roughly five times faster than V7, per an AIWorthIt review that tracked the July 24, 2026 rollout. That speed is how an estimated 1.2 to 2.5 million daily active users, out of the roughly 20 million registered accounts on the platform, keep getting grids back without long queues most of the day. Within 30 to 90 seconds, depending on server load and whether you’re on a fast or relaxed job, Midjourney returns a grid of four separate images. Expect four different takes on the same fox-in-snow concept: different head angles, different light falloff, different framing, since the model treats every phrase as a soft suggestion rather than an exact instruction. That’s normal, and it’s the reason the next few steps focus on narrowing the range of outcomes rather than fighting it.

Under the grid you’ll see two rows of buttons. U1 through U4 upscale one of the four images into a larger, sharper final version. V1 through V4 generate four new variations based on one of the four originals, keeping its rough composition but changing the details. In the web app, these appear as icons you hover over rather than labeled buttons, but they do the same job. Pick the image closest to what you wanted, upscale it, and you have your first finished Midjourney output.

With the fox prompt above, expect a wide 16:9 frame with the fox roughly centered or slightly off to one side, snow-covered branches framing the shot, and a warm golden-hour color cast across the whole scene. The –stylize 250 value pushes the render toward a more polished, deliberately composed look rather than a flat, literal illustration of the words. If your grid instead comes back looking washed out or oddly cropped, that’s usually a sign to revisit the parameters in Step 5 before assuming the prompt itself failed.

Step 4: Write Prompts That Work With the Six-Part Formula

Most disappointing Midjourney results trace back to underspecified prompts, not a limitation of the model. A reliable fix is to build every prompt from the same six ingredients, in roughly this order, since Midjourney weighs earlier words more heavily than later ones.

Part	Question It Answers	Example
Subject	Who or what is the focus?	a barista
Action	What are they doing?	pouring latte art into a ceramic cup
Environment	Where is this happening?	inside a small brick-walled coffee shop
Lighting	What’s the light source and mood?	warm morning light through a front window
Lens/camera	How was it “shot”?	85mm lens, shallow depth of field
Style	What’s the overall aesthetic?	editorial photography, muted color grade

Stitched together, that becomes: “a barista pouring latte art into a ceramic cup, inside a small brick-walled coffee shop, warm morning light through a front window, 85mm lens, shallow depth of field, editorial photography, muted color grade.” Every clause earns its place. Compare that with a vague prompt like “a coffee shop photo,” which leaves Midjourney to guess at framing, lighting, and mood, usually landing somewhere generic.

You don’t need all six parts every time. Quick concept sketches can skip lens and lighting entirely. But when a result feels flat or off-brief, work through the list and ask which piece you left out. That single habit fixes more bad generations than any parameter covered in the next section.

Step 5: Control Composition with Core Parameters

Parameters are the flags you add to the end of a prompt, each starting with two dashes, and they give you direct control instead of hoping the model reads your intent correctly. These seven cover most day-to-day needs.

Parameter	What It Does	Example
–ar	Sets the aspect ratio (width:height)	–ar 16:9
–stylize or –s	Controls how strongly Midjourney’s house aesthetic is applied, from 0 to 1000	–stylize 100
–chaos or –c	Increases variation between the four grid results, from 0 to 100	–chaos 30
–weird or –w	Pushes results toward unusual, experimental compositions	–weird 250
–no	Excludes a specific element from the image	–no text, watermark
–tile	Generates an image whose edges line up when repeated side by side, useful for patterns and textures	–tile
–v 7	Locks the generation to a specific model version	–v 7

Low –stylize values (0 to 100) keep Midjourney close to your literal words, which helps for technical or product-accurate work. Higher values (500 and up) let the model take more artistic liberty, often at the cost of prompt accuracy. –chaos works similarly but affects variation between the four grid images rather than adherence to your text, so a low –chaos value gives you four close variations of the same idea, while a high one gives you four genuinely different interpretations worth comparing.

Stack parameters at the end of the prompt, after your six-part description, separated by spaces:

/imagine prompt: a barista pouring latte art into a ceramic cup, inside a small brick-walled coffee shop, warm morning light through a front window, 85mm lens, editorial photography --ar 4:5 --stylize 150 --chaos 10 --no text --v 7

Change one parameter at a time while you’re learning. Adjusting –ar, –stylize, and –chaos together in a single prompt makes it nearly impossible to tell which change actually improved, or hurt, the result.

Step 6: Lock In a Consistent Look with Style and Character References

Once you’ve generated an image whose look you want to reuse, style references let you carry that aesthetic into new prompts instead of re-describing it from scratch every time. Upload or link a reference image and add –sref followed by its URL:

/imagine prompt: a woman walking through a busy Tokyo street at night, neon signs reflecting on wet pavement --sref [reference-image-url] --sw 250 --ar 3:2 --v 7

The –sw flag sets style weight, how strongly the reference influences the output, typically somewhere between 0 and 1000. Push it higher when you want near-identical color grading and rendering style across a whole set of images, which is exactly the technique used in the branded project near the end of this tutorial.

Character consistency works the same way but targets a person or figure instead of a general aesthetic. Use –cref (character reference) with a clear, front-facing reference image, plus –cw to control how strongly facial features and proportions carry over:

/imagine prompt: the same woman now sitting at a ramen counter, warm interior lighting --cref [reference-image-url] --cw 100 --ar 3:2 --v 7

Omni-Reference, a broader version of the same idea that can lock in a specific object, product, or non-human character across a series rather than just a human face, actually arrived earlier than most people assume: Midjourney built it into the V7 launch on April 3, 2025, alongside Draft Mode, rather than saving it for a later model line. It runs on its own –oref and –ow flags, which control reference strength the same way –cref and –cw do for characters, a syntax Midjourney and ThatPainter have documented since 2025. If you’re building anything with a recurring mascot, product shot, or brand character, Omni-Reference is worth testing before you fall back to writing longer and longer text descriptions to force consistency.

Run the ramen counter prompt above with a strong –cref image, and the payoff is obvious the moment the grid loads: the same face, hairstyle, and rough build carry over into a completely different setting and pose. Without –cref, asking for “the same woman” a second time produces someone who merely resembles the first image, not a match. That gap is exactly what reference parameters exist to close, and it’s the difference between a one-off image and a usable recurring character.

Step 7: Upscale, Vary, Edit, and Retexture Your Images

A grid result isn’t the finished product, it’s a rough draft. Once you upscale an image (Step 3), Midjourney’s editor unlocks a second layer of tools that work on the upscaled version directly, rather than generating a fresh grid.

• Vary (Subtle): generates small variations while keeping composition, subject, and framing nearly identical, good for picking between near-final options
• Vary (Strong): keeps the general concept but allows bigger changes to pose, background, and detail
• Editor: lets you select a region of the image and regenerate just that area, useful for removing an unwanted object or fixing a warped hand without redoing the whole image
• Retexture: keeps the composition and shapes in your image fixed while changing surface materials, useful for testing a product mockup in different finishes without re-prompting from zero
• Zoom Out: extends the canvas beyond the original frame, revealing more of the scene while keeping the center intact

Use Editor before you consider a prompt a failure. A single distracting element (an extra finger, a stray object in the background, a logo-like shape Midjourney invented) is often a 30-second regional fix rather than a reason to start over with a new prompt.

Step 8: Turn On Stealth Mode for Private Generations

By default, every image you generate is visible in Midjourney’s public feed and searchable by anyone, including your prompt text. That’s fine for personal projects, but it’s a problem if you’re prototyping client work, an unannounced product, or anything under an NDA.

Stealth Mode fixes this by keeping your generations private, visible only to you, but it’s restricted to the Pro and Mega plans. Basic and Standard subscribers can’t turn it on at any price within those tiers, so if privacy matters for your work, factor that into the plan you chose back in Step 1. To enable it, open your account settings in the web app and toggle Stealth Mode on. It applies going forward, not retroactively, so anything you generated before switching it on stays in the public feed.

Step 9: Know Your Commercial Usage Rights

Paid Midjourney subscribers generally receive commercial usage rights to the images they generate, meaning you can sell prints, use outputs in client work, or publish them commercially without paying an extra licensing fee. There’s one revenue-based catch worth knowing before you build a business around it: companies or individuals with more than $1 million in annual gross revenue are required to subscribe to Pro or Mega to retain those commercial rights, according to Midjourney’s terms as reported across multiple 2025 and 2026 pricing breakdowns. Basic and Standard don’t carry commercial rights once that revenue threshold is crossed.

Ownership and copyright of AI-generated images is a separate, murkier question from usage rights, and it’s still unsettled at a legal level. The U.S. Copyright Office’s guidance on AI and copyright has generally held that works lacking sufficient human authorship don’t qualify for copyright protection on their own, and the USPTO’s AI initiative tracks how that standard applies to patents and related intellectual property questions. None of that stops you from using Midjourney images commercially under your subscription’s terms, but it does mean you may have limited ability to stop someone else from using a very similar AI-generated image. If your use case depends on exclusive ownership, that’s worth a conversation with an actual IP attorney rather than a tutorial.

In practice, most people run into this question when a client asks whether a logo or hero image “belongs” to them outright. The honest answer is that your subscription grants usage rights, not a guarantee of exclusivity. Two businesses could theoretically end up with strikingly similar AI-generated marks through unrelated prompting, and neither would have a strong copyright claim against the other. For low-stakes work like blog headers or internal decks, that’s rarely a practical problem. For a primary brand logo or packaging, it’s worth pairing an AI-generated starting point with enough human editing to strengthen any later ownership claim.

Step 10: Organize and Automate Your Downloads

Midjourney does not offer an official public API. If you see a product marketed as a “Midjourney API,” it’s an unofficial third-party wrapper that automates the Discord bot or web app behind the scenes, not something published or supported by Midjourney itself. Using one risks violating Midjourney’s terms of service and can get an account suspended, so treat any service promising programmatic access with caution and read its fine print before connecting your account to it.

What you can safely automate is the boring part that happens after generation: sorting the files you’ve already downloaded. Export your images from the web app’s library (select an image, then Save or Download), and once you’ve got a folder full of PNGs, a short script keeps them organized by date without manual dragging and dropping:

#!/bin/bash
# Sort downloaded Midjourney images into per-day folders
SRC="$HOME/Downloads"
DEST="$HOME/Midjourney-Projects"
mkdir -p "$DEST"

for file in "$SRC"/*.png; do
  [ -e "$file" ] || continue
  day=$(date -r "$file" +%Y-%m-%d)
  mkdir -p "$DEST/$day"
  mv "$file" "$DEST/$day/"
done

echo "Sorted $(find "$DEST" -name '*.png' | wc -l) images into $DEST"

Save that as sort-midjourney.sh, run chmod +x sort-midjourney.sh once, then run ./sort-midjourney.sh any time your downloads folder gets cluttered. It’s a small thing, but once you’re generating dozens of images a week for the kind of project in the next section, a flat unsorted downloads folder turns into its own problem.

Complete Project: Build a Six-Image Branded Header Set

Here’s where the individual steps come together. The goal: six blog header images for a fictional coffee brand, all sharing a consistent style, color palette, and aspect ratio, ready to drop into a CMS without further editing. This is the exact workflow to reuse for any project that needs a matching set rather than one-off images.

First, generate one image that nails the look you want, using the six-part formula from Step 4. Once you’re happy with it, upscale it and save its job URL, since you’ll reuse it as a style reference for the rest of the set.

/imagine prompt: minimalist flat-lay of a laptop and coffee cup on a wooden desk, soft morning light from the left, editorial product photography, muted warm tones --ar 16:9 --stylize 150 --v 7

Once that first image is upscaled, grab its URL and reuse it as a style reference in the remaining five prompts, changing only the subject each time and keeping the aspect ratio, stylize value, and reference identical:

/imagine prompt: minimalist flat-lay of an open notebook and pen on a wooden desk, soft morning light from the left, editorial product photography, muted warm tones --sref [image-1-url] --ar 16:9 --stylize 150 --v 7

/imagine prompt: minimalist flat-lay of a croissant on a ceramic plate on a wooden desk, soft morning light from the left, editorial product photography, muted warm tones --sref [image-1-url] --ar 16:9 --stylize 150 --v 7

/imagine prompt: minimalist flat-lay of a potted plant and reading glasses on a wooden desk, soft morning light from the left, editorial product photography, muted warm tones --sref [image-1-url] --ar 16:9 --stylize 150 --v 7

/imagine prompt: minimalist flat-lay of a phone showing a calendar app and a pen on a wooden desk, soft morning light from the left, editorial product photography, muted warm tones --sref [image-1-url] --ar 16:9 --stylize 150 --v 7

/imagine prompt: minimalist flat-lay of a stack of books and reading glasses on a wooden desk, soft morning light from the left, editorial product photography, muted warm tones --sref [image-1-url] --ar 16:9 --stylize 150 --v 7

Because the aspect ratio, style reference, and stylize value stay locked across all six prompts, the only thing changing between generations is the subject itself. That’s what makes the final set look like it came from one photoshoot rather than six unrelated experiments. Upscale your favorite from each grid, run them all through the sorting script from Step 10, and you have a complete, on-brand header set ready to publish.

Line the six finished images up side by side and the effect should be immediate: the same wooden desk tone, the same soft directional light, the same restrained color palette, even though the subject changes from a laptop to a croissant to a stack of books. If one image in the set looks noticeably off (colder lighting, a different desk material, a busier composition), it’s almost always because –sw wasn’t strong enough relative to how detailed that particular subject’s prompt was. Bump the style weight on the outlier and regenerate just that one prompt rather than redoing the whole set.

Advanced Tips for Power Users

• Multi-prompts: separate two ideas with a double colon (fox :: forest) to have Midjourney weigh them as distinct concepts instead of blending them into one phrase, and add a number after each (fox::2 :: forest::1) to control the balance
• Permutation prompts: wrap comma-separated options in curly braces, like a {red, blue, green} sports car, and Midjourney queues one separate job per option automatically instead of you re-typing the prompt three times
• –seed: every generation uses a random starting seed unless you fix one with –seed [number], which lets you reproduce the same base image later while tweaking other parameters, useful for isolating exactly what a single parameter change does
• Personalization: Midjourney’s 2026 model line supports a personalization mode that leans on images you’ve previously liked to bias new generations toward your own taste, handy once you’ve built up a history on the platform
• Draft Mode: renders rough concepts faster and cheaper than a standard generation, which makes it the right tool for exploring ten different compositions before you commit fast hours to polishing the one that actually works
• /describe: upload an existing image and Midjourney suggests text prompts that could plausibly have produced something similar, a fast way to reverse-engineer a style you found elsewhere into language you can then adapt and reuse

Midjourney vs Nano Banana Pro vs Adobe Firefly vs Stable Diffusion in 2026

Midjourney isn’t the only serious option in 2026, and depending on your budget and workflow, it may not even be the best fit. Independent AI-tool pricing trackers put the major options side by side roughly as follows, though exact numbers shift often enough that it’s worth double-checking each vendor’s current pricing page before you commit to a plan.

Tool	Starting Price	Max Resolution	Typical Speed	Access
Midjourney	$10/mo	HD / 2K	~30-90s	Web app + Discord
Nano Banana Pro (Google)	$0.15-$0.24/image	3840×2160	Varies by request	Gemini app / API
Adobe Firefly (Image 5)	$19.99/mo	4MP	~12s	Web + Creative Cloud
Stable Diffusion 3.5	Free (self-host)	1024×1024	~30s (local GPU)	Local install

Midjourney’s edge is prompt control and a large, well-documented parameter set built up over years, plus a community feed you can learn from by studying other people’s prompts. Adobe Firefly integrates directly into Photoshop and Illustrator, which matters if your output needs to slot into an existing Creative Cloud workflow. Google’s Nano Banana Pro, which we cover in a separate step-by-step tutorial, charges per image rather than per month and currently tops out at a higher native resolution. Stable Diffusion, walked through in our FLUX and ComfyUI guide and our earlier ComfyUI workflow tutorial, is the only option here with zero ongoing cost, provided you already own a capable GPU or are willing to invest in dedicated local AI hardware. For a wider field of tools beyond these four, our AI image generator comparison covers the full 2026 lineup.

Adobe’s Firefly is trained on licensed Adobe Stock content, which some commercial teams prefer specifically because it sidesteps the ownership questions raised in Step 9. If your organization has a strict policy against using tools trained on scraped web images, that’s a real reason to pick Firefly over Midjourney regardless of price.

As a rule of thumb: pick Midjourney when prompt control and a mature parameter set matter more than raw resolution, pick Nano Banana Pro when you need the highest native resolution and want to pay only for what you generate, pick Firefly when the output has to live inside an existing Adobe workflow, and pick Stable Diffusion when the budget is zero and you already have the hardware to run it. None of these choices are permanent. Plenty of working creators keep two subscriptions running at once and switch depending on the job.

Common Pitfalls and Troubleshooting

Six Pitfalls That Waste Your Fast Hours

1. Overstuffed prompts. Piling on ten competing style words (“photorealistic, anime, oil painting, 3D render”) confuses the model instead of refining it. Pick one aesthetic direction per prompt.
2. No aspect ratio specified. Midjourney defaults to a square image if you skip –ar, which is rarely what you need for a header, a phone wallpaper, or a print piece. Set it explicitly every time.
3. Vague negative prompts. –no bad quality doesn’t work well because “bad quality” isn’t a concrete visual object. –no text, watermark, blur works because those are things the model can actually identify and remove.
4. Changing too many variables at once. Adjusting –stylize, –chaos, and your wording in the same prompt makes it impossible to know which change caused the result you’re looking at.
5. Expecting legible text. Image models, Midjourney included, still struggle to render accurate typography inside an image. Add text in a separate design tool afterward instead of prompting for it directly.
6. Burning fast hours during early exploration. Rough out composition ideas in Relax Mode first, then switch to fast generations only once you’re polishing a nearly-final concept.

Nine Common Problems and How to Fix Them

• The image doesn’t match my prompt at all. Shorten the prompt and move your most important concept to the front. Midjourney weighs earlier words more heavily than later ones.
• My prompt got blocked by moderation. Check for flagged terms: graphic violence, real public figures by name, certain trademarked characters, and some brand names all trigger automatic blocks. Rephrase around the specific word, not just the general idea.
• The upscale button is missing or grayed out. You’re likely out of fast hours for the month. Switch to Relax Mode if your plan includes it, or wait for your monthly reset.
• My job has been stuck in the queue for a long time. High-traffic hours (evenings and weekends in US time zones) slow Relax Mode considerably. Switch to a fast generation if the wait is a problem, or try again later.
• –sref isn’t matching the reference style closely enough. Raise –sw (style weight) toward 500 or higher, and make sure your reference image itself has a clean, distinct style rather than a busy or mixed one.
• My character’s face keeps changing between generations. Use a clear, front-facing –cref image and raise –cw (character weight). A blurry or angled reference photo produces inconsistent results no matter how high you set the weight.
• I keep getting nearly identical results. Raise –chaos to push more variation between the four images in a single grid.
• The Discord bot isn’t responding to my commands. Check Midjourney’s status channel for an ongoing outage, and confirm your Discord account is still connected under your Midjourney account settings.
• I ran out of fast hours mid-month. Switch to Relax Mode if your plan supports it, or upgrade to the next tier from your account settings. Changes apply immediately and you keep every image you’ve already generated.

Frequently Asked Questions

Is Midjourney still Discord-only in 2026?

No. The web app at midjourney.com is now the primary way most people use Midjourney, and Discord is an optional sign-in and generation method rather than a requirement.

How much does Midjourney cost per month?

Plans run from $10 a month on Basic up to $120 a month on Mega, with Standard ($30/mo) and Pro ($60/mo) in between. Paying annually cuts the effective monthly cost by about 20 percent on every tier.

Can I use Midjourney for free?

Not on the main web app or Discord. New accounts require a paid subscription starting at $10 a month. The only free trial Midjourney currently offers lives in its separate niji・journey mobile app, which is a different, more limited product.

What’s the difference between –stylize and –chaos?

–stylize controls how much artistic license Midjourney takes with your literal wording: low values stay close to your text, high values look more “designed.” –chaos controls how different the four images in a single grid are from each other, independent of how closely any of them match your prompt.

Can I use Midjourney images commercially?

Yes, if you’re on a paid plan. The one exception: organizations with more than $1 million in annual gross revenue need to be on the Pro or Mega plan specifically to retain commercial usage rights, according to Midjourney’s terms. Always check the current terms of service before building a business around any specific output.

Why do my Midjourney images not match my prompt?

Usually it’s an overstuffed prompt, a missing aspect ratio, or a –stylize value set too high for how literal you need the result to be. Work through the six-part formula in Step 4 and check which element you left out before assuming the model simply got it wrong.

Is there an official Midjourney API?

No. Midjourney has no official public API as of 2026. Any product marketed as a “Midjourney API” is an unofficial third-party wrapper, and connecting one to your account carries a real risk of violating Midjourney’s terms of service.

What happened to older Midjourney versions like V6?

Midjourney has moved its default forward with each major release. V7 launched on April 3, 2025, adding Draft Mode and Omni Reference, according to Midjourney’s own documentation, and became the documented default model on June 17, 2025, replacing V6.1, per Midjourney, then gave way to V8.1, which Midjourney first shipped as an alpha in April 2026 before setting it as the documented default on June 10, 2026, reported VidMuse AI, with improved prompt handling, text, and output speed fast enough to render a standard-definition image in about 4 seconds and an HD image in about 12, before V8.1 was itself replaced by V8.2 on July 24, 2026, according to a Blake Crosley version-history guide, bringing native 2K output and a rebuilt personalization system with it. If you don’t specify –v in your prompt, you’ll get whichever version is currently set as your account’s default, so add –v 8.2 explicitly if you want to guarantee consistency across a project.

For background on the company itself, including its funding history and product timeline, Midjourney’s Wikipedia entry tracks the platform’s development since its 2022 public launch. And if you’re comparing image tools more broadly, Adobe’s Firefly product page and Stability AI’s newsroom are worth bookmarking alongside Midjourney’s own release notes, since all three ship model updates on their own schedules throughout the year.