How to Use Midjourney: A Beginner’s Tutorial for 2026 (V8.2)

This tutorial takes you from signing up to your first Midjourney image, edited, upscaled and downloaded, in six steps on midjourney.com. You need a Google or Discord account and a paid plan: there is no free trial, and the cheapest plan, Basic, costs $10 a month before tax (Midjourney plans, checked Sept. 25, 2026). Every step follows Midjourney’s official documentation as of September 25, 2026.

A lot has changed since the February 2025 version of this guide: the default model has been V8.2 since July 24, 2026, the website has replaced Discord as the main interface, the --q parameter no longer works with the latest models, and a new Edit model changes images from plain written instructions. The old screenshots, which showed an interface that no longer exists, have been removed. For full pricing, video features and usage rights, see our complete Midjourney guide.

What You Need Before You Start

Sources: plan comparison, free trials, Stealth Mode, checked Sept. 25, 2026. Plans are billed in U.S. dollars; sales tax is added at checkout based on your billing address.

How to Use Midjourney in 6 Steps

1. Create an account and pick a plan.
2. Type your first prompt in the Imagine bar.
3. Set the shape and style with parameters.
4. Lock in a look with a Style Reference or a Moodboard.
5. Refine the image: variations, Edit, upscaling.
6. Download and organize your images.

Each step is detailed below. Button names come from the official documentation; Midjourney is also testing a redesigned interface on alpha.midjourney.com, so a few buttons may move.

Step 1: Create an Account and Subscribe

1. On midjourney.com, click Sign Up.
2. Choose Continue with Google or Continue with Discord. If you have ever made images in Discord before, sign in with Discord first: a new subscription started with Google can’t be linked to your old image history later (Logging In & Connecting Accounts).
3. Open your Manage Subscription page and scroll down to the plans.
4. Heads up: yearly billing is selected by default, with the full year paid upfront. Click the Monthly Billing tab if you’d rather pay month to month (How to Subscribe).
5. Click Subscribe under the plan you want, then pay through the Stripe checkout.

Which plan to start with (monthly prices before tax, checked Sept. 25, 2026):

To try the tool, one month of Basic is enough: a batch of four standard-definition images costs 0.8 minutes of GPU time in V8.2 (Version page), so the 200 included minutes cover roughly 250 prompts. Standard makes sense once you generate regularly, because Relax Mode doesn’t use up Fast time.

Step 2: Type Your First Prompt in the Imagine Bar

1. Open the Create page on midjourney.com.
2. Type your description, the “prompt,” into the Imagine bar at the top of the page.
3. Press Enter (or the send button on mobile). Midjourney generates four images, which you can watch render until they reach 100%.

Tip: Ctrl + Enter (Cmd + Enter on a Mac) runs the prompt and keeps the text in the Imagine bar, so you can tweak it and run it again (Creating on Web).

How to Write a Good Midjourney Prompt

The documentation recommends short, descriptive prompts over long lists of instructions. Describe what you want to see, not what Midjourney should do, and spell out the details that matter to you:

• Subject: who or what (person, animal, object, place);
• Medium: photo, painting, illustration, sculpture, sketch;
• Environment: indoors, forest, city, underwater;
• Lighting: soft, neon, studio lights, overcast;
• Color: vibrant, pastel, monochromatic, black and white;
• Mood: calm, playful, gloomy, energetic;
• Composition: portrait, close-up, bird’s-eye view.

Use exact numbers (“three cats” rather than “cats”) and precise words (“enormous” rather than “big”). For text inside the image, wrap it in double quotation marks: V8 renders it better than earlier versions (V8 announcement, Text Generation).

Copy-and-Paste Prompt Examples

All of these prompts are ready for V8.2. Paste them as is, then swap in your own subject.

Realistic photo:

Close-up photo of an elderly fisherman mending a net on a wooden dock at sunrise, soft golden light, shallow depth of field --ar 3:2 --raw

Illustration for a blog post or slide deck:

Flat vector illustration of a cozy reading nook with a cat asleep on a pile of books, pastel colors, white background --ar 4:5

Landscape in banner format:

Wide shot of a misty pine forest in the Rocky Mountains at dawn, cinematic lighting, muted greens and blues --ar 16:9 --s 250

Poster with text:

Vintage art deco travel poster of Route 66 with the words "ROUTE 66" in bold letters --ar 2:3

Product shot with no stray logos:

Studio product photo of a matte ceramic coffee mug on a concrete table, soft shadows, minimal background --ar 1:1 --no text, logo

Step 3: Set the Shape and Style With Parameters

Parameters go at the end of your prompt, after a space, with no commas or periods. vibrant poppies --ar 2:3 works; vibrant poppies--ar 2:3, - - ar and --ar 2:3, don’t (Parameter List).

Sources: Aspect Ratio, Stylize, Chaos / Variety, Raw, Version, checked Sept. 25, 2026. In HD, the widest supported aspect ratio is 4:1 (14:1 in SD).

Instead of typing the same parameters every time, click the settings icon in the Imagine bar: aspect ratio, SD or HD, version, Raw, stylization, weirdness, variety, GPU speed and Stealth Mode can all be set as defaults for every prompt.

Step 4: Lock In a Look With Style References and Moodboards

Style Reference (--sref)

A Style Reference captures the look of an image (colors, medium, textures, lighting) without copying what’s in it.

1. Click the image icon in the Imagine bar to open the images panel.
2. Upload your image (10 MB max) or pick one you’ve already uploaded.
3. Drag it into the Style Reference section.
4. Write a prompt that describes the content you want, not an instruction: “ballpoint pen sketch of a bunny” rather than “copy this style and make a bunny.”

You can also use a style code: --sref random picks one at random, and the Styles tab on the Explore page lets you browse codes, try them (Try Style) and add them to your prompt (Copy). --sw (0 to 1000, default 100) sets how strongly the style applies.

Watercolor portrait of a red fox wearing a knitted scarf --sref random

Moodboards and Personalization (--p)

A Moodboard is a set of images that defines a broader style than a single reference. On the Moodboards page, click New Moodboard, name it and add images (uploaded, by URL or from your gallery). To use it, turn on the Personalization button next to the Imagine bar and select it, or paste its ID at the end of your prompt (--p followed by the ID).

A Personalization profile learns your taste: on the Personalize page, click your Global V7 Profile and pick your favorite images from a grid until it unlocks. That profile works with V8.2. In both cases, --s controls how much of your style shows up in the image.

Step 5: Refine the Image With Variations, Edit and Upscale

Click an image to open it full-screen: the tools live in the Creation Actions section. If a button is missing, click More options and check it (Modifying Your Creations).

• Vary (Subtle / Strong): four new versions of the image, close to it or clearly different.
• Remix (Subtle / Strong): same idea, but the prompt goes back into the Imagine bar so you can change it before rerunning.
• Upscale (Subtle / Creative): doubles the size of an SD image (1,024 → 2,048 px at 1:1). Subtle keeps the image as is; Creative adds detail and can fix an awkward hand. Upscaling can cost twice the GPU time of the original generation, and HD images can’t be upscaled further (Upscalers).

To redo an SD batch in 2K, hover over the prompt in your Create feed and click Run batch as HD: the same prompt reruns in HD with the same seed.

Edit Images With the Edit Model

The Edit model, available in V8.1 and V8.2, replaces Omni Reference, Character Reference and Retexture. You can use it three ways:

• Quick Edit puts the image in the Imagine bar: type an instruction (here, direct instructions are fine).
• Open Editor loads the Editor: select an area with the Erase brush or Smart Select, describe what should appear, then click Submit Edit. Since Sept. 24, 2026, only the selected pixels change (announcement).
• Up to 4 reference images dropped into the Attach to prompt section: today’s way to keep the same character or object from one image to the next.

Example Quick Edit instructions:

make it night time with warm street lights

let's see this character from behind

hand drawn pencil sketch of this image

The Edit model matches the aspect ratio of the first attached image; add --ar to change it. --raw makes results follow your instruction more closely. It doesn’t work with --tile, and its results can’t be used with Remix.

Animate Your Image

The Animate buttons turn the image into a 5-second video, either automatically (Auto) or with your own motion prompt (Manual). Length, resolution and cost are covered in our Midjourney guide.

Step 6: Download and Organize Your Images

• One image: open it full-screen and click the download icon near the prompt.
• Several images: on the Organize page, select them (click and drag, or Shift + click), then download them together. Whole folders can be downloaded too (Organizing Your Creations).
• Folders: the folder icon next to the Imagine bar creates folders you can generate directly into.

On Basic and Standard, everything you make is public on Midjourney’s website and can show up on the Explore page. Trashing an image doesn’t make it private: only Stealth Mode on Pro and Mega does.

Work Faster With Draft Mode and Conversational Mode

Draft Mode is for exploring ideas cheaply. Turn it on with the lightning-bolt button next to the Imagine bar, or add --draft to your prompt. In V8.2, one Draft prompt returns 24 sketches at 512 × 512 pixels for 0.4 minutes of GPU time. Then click Vary or Remix on the sketch you like to regenerate it at normal resolution (SD or HD, per your settings). Draft Mode for V8.2 is website-only; it isn’t available in Discord.

Pair it with --sref random to get 24 images in 24 different styles, a quick way to find an art direction (announcement):

Isometric illustration of a tiny bakery on a street corner --draft --sref random

Conversational Mode (its own button in the Imagine bar) lets you describe your idea in everyday language, typed or spoken, and an AI writes the prompts for you. It works in other languages too. To follow up, refer to images by number: “run a variation of image 4.” Voice input requires Draft Mode to be on.

Using Midjourney in Discord

Discord is still an option, on the same subscription (Discord Quick Start):

1. Join the Midjourney server (the + button at the bottom of your server list, then Join a Server).
2. Open a general-# channel or newbie-1 through newbie-10 (higher newbie channels are archived), or send the bot a direct message.
3. Type /imagine, then your prompt, and accept the Terms of Service the first time.
4. Under the grid, U1–U4 separate an image without enlarging it and V1–V4 make variations; once an image is separated, Vary (Strong/Subtle) and Upscale (Subtle/Creative) appear.

Several tools are missing in Discord: the Editor, V8.2 Draft Mode, Conversational Mode, and creating Moodboards or Personalization profiles. Images made in public channels are visible to everyone, even in Stealth Mode. For beginners, the website is simpler.

Common Midjourney Mistakes and How to Fix Them

Paying for a Full Year by Accident

The subscription page defaults to yearly billing. Click Monthly Billing before Subscribe to pay month to month.

Losing Your Old Images

If you used Midjourney in Discord before, sign in with Discord, not Google. A Google account with a subscription and a Discord account that already has images can’t be merged.

Adding --q 2 for “Better Quality”

The quality parameter isn’t compatible with V8.1 and V8.2, according to the compatibility chart. For a bigger image, use --hd or the Upscale buttons.

Writing “Without X” in the Prompt

“A party with no cake” may well produce a cake. Describe what you want and put the exclusion in a parameter: --no cake. Note that --no reads each word separately: --no modern clothing is read as “no modern” and “no clothing,” which can trigger moderation.

A Parameter Is Ignored or Throws an Error

Check for the space before the two dashes, remove any punctuation and keep parameters at the very end. --ar doesn’t accept decimals: write --ar 139:100, not --ar 1.39:1.

Misspelled Text in the Image

Use double quotes (not single), keep the words short, add “with the words” or “written,” then try --raw or a lower --s value. The Editor can fix a stray letter.

A Prompt That Goes Off the Rails

Long prompts full of directions (“make them bright, draw them with…”) muddy the result. Go back to a short description: subject, medium, lighting, framing.

Images Too Close to Famous Characters

Copyrighted characters, logos and brands are a legal risk, especially for commercial use: Disney, Universal and Warner Bros. are suing Midjourney in U.S. court over exactly this (TechCrunch, July 2026). Our article on Studio Ghibli–style AI images covers the debate.

Midjourney Alternatives

Midjourney has no free tier. To make images without paying, the Google Gemini app includes image generation with Nano Banana 2 even without a subscription (Gemini Apps features and limits, checked Sept. 25, 2026). ChatGPT also creates and edits images inside a conversation (see our guide to image generation in ChatGPT). Midjourney’s edge is style control: references, Moodboards and Personalization.

FAQ

Is Midjourney free?

No. There’s no free plan or trial on midjourney.com or Discord; only the niji・journey mobile app offers a limited trial. The Basic plan costs $10 a month before tax (checked Sept. 25, 2026).

Do you still need Discord to use Midjourney?

No. Everything works on midjourney.com with a Google or Discord login. The website even has tools Discord lacks: the Editor, V8.2 Draft Mode, Conversational Mode and Moodboards.

How many images do you get with the Basic plan?

The 200 minutes of Fast GPU time work out to about 250 prompts of four standard-definition images each in V8.2 (0.8 minutes per prompt), fewer in HD (1.3 minutes) or if you upscale.

How do you keep the same character across images?

Use the Edit model: drag up to four images of your character into the “Attach to prompt” section of the Imagine bar, then describe the new scene. It replaces Character Reference and Omni Reference as of V8.2.

Can you use Midjourney images commercially?

Yes, on every paid plan, as long as you avoid copyrighted works, characters and brands. A company with more than $1 million in annual gross revenue must be on the Pro or Mega plan (Terms of Service).

Which Midjourney version is the default?

V8.2, since July 24, 2026. Add --v 7 to the end of a prompt to use the previous model, or change your default version in the Imagine bar settings.