How to Use Midjourney: The Beginner’s Guide (First Prompt, V8.2, and a Clean Upscale)
CodexFuse Team · 2026-10-28T10:34:36.558Z · 16 min read
Learn how to use Midjourney in 2026: subscribe for $10, write your first prompt on the web, upscale and edit with V8.2. No free trial, no Discord required.
Midjourney turns a sentence into an image. You describe what you want, it returns four versions in seconds, and you refine the one you like until it is ready to use. Learning how to use Midjourney takes about an hour: subscribe to a plan, write your first prompt on the website, and learn the handful of buttons that turn a rough grid into a finished image.
This guide is current for October 2026. It teaches the web interface first, because that is now the main way in, and it covers the Discord route second. It also corrects the two things most older guides still get wrong: there is no free trial, and the default model is now V8.2, not V8.1. If you are comparing image generators before you commit, our best AI image generator guide puts Midjourney next to the other 2026 options by job.
How do you use Midjourney for the first time?
You use Midjourney for the first time by signing up with a Google or Discord account, subscribing to a paid plan, and typing a description at midjourney.com/imagine. Midjourney returns four images in a grid, and you pick one to upscale, vary, or edit.
- Create an account at midjourney.com. Sign in with Google or Discord. There is no email-and-password option, and billing runs through Stripe.
- Subscribe to a plan. Generation needs a paid plan. Basic costs $10 a month and is enough to learn on.
- Open midjourney.com/imagine. This is the web prompt bar. Type a description and press Enter.
- Wait for the four-image grid. A standard V8.2 image takes about 4 seconds. You get four variations of your prompt.
- Click the image you like best. Choose Upscale for a single high-resolution version, or Vary for four new images like it.
- Download it. On the web, each image has a download button. Your history lives in the gallery, searchable by keyword.
That is the whole loop. Everything else in this guide is about spending less of your monthly GPU time to get there.
What is Midjourney, and what is it good at?
Midjourney is an AI image generator that creates images from text descriptions called prompts. You describe what you want in plain language, it generates four options, and you refine from there. It is a creative collaborator, not a vending machine for exact images: the first generation is rarely the final result, and good Midjourney work usually takes 5 to 15 generations.
It is genuinely good at stylized art, concept imagery, and anything where mood matters more than precision: portraits, landscapes, brand moodboards, fantasy illustration, product concept art, and social media graphics. It is weaker at text inside images, specific real people, precise spatial arrangements, and technical diagrams. If you need a word spelled correctly inside an image, put the text in double quotes in your prompt: V8.2 handles quoted text substantially better than earlier versions, but the safe workflow is still to generate without text and add it in Canva, Photoshop, or Figma afterward.
Midjourney is a small, self-funded company that has been profitable since 2022, and its Discord community has roughly 18 to 21 million members. Those members are not all paying subscribers. What matters for a beginner is simpler: the tool is mature, the web app is stable, and the beginner path no longer runs through Discord chaos unless you want it to.
How much does Midjourney cost in 2026?
Midjourney costs $10 to $120 a month across four plans, with annual billing taking 20% off every tier. There is no free plan, and there is no free trial on the website or in Discord. The only trial anywhere is a limited one inside the separate niji journey mobile app for iOS and Android.
| Plan | Monthly | Annual (per month) | Fast GPU time | Relax mode | Stealth mode |
|---|---|---|---|---|---|
| Basic | $10 | $8 | 3.3 hours (200 minutes) | No | No |
| Standard | $30 | $24 | 15 hours | Unlimited images | No |
| Pro | $60 | $48 | 30 hours | Unlimited images | Yes |
| Mega | $120 | $96 | 60 hours | Unlimited images | Yes |
Midjourney sells GPU compute time, not image slots. A standard-definition image prompt costs about 0.8 GPU minutes, and an HD prompt costs about 1.3 GPU minutes. Extra fast time costs $4 per hour on any plan. Fast hours reset each billing cycle and do not roll over.
The table below is our own computation from those published GPU-minute costs. It shows what each plan actually buys in a month, so you can pick a plan by output instead of by price alone.
| Plan | Standard prompts per month | HD prompts per month | Draft prompts per month | Cost per GPU hour |
|---|---|---|---|---|
| Basic ($10) | ~250 | ~150 | ~500 | $3.03 |
| Standard ($30) | ~1,125 | ~692 | ~2,250 | $2.00 |
| Pro ($60) | ~2,250 | ~1,384 | ~4,500 | $2.00 |
| Mega ($120) | ~4,500 | ~2,769 | ~9,000 | $2.00 |
Basic runs 3 concurrent fast jobs; Standard also runs 3, while Pro and Mega run 12. Standard is the value break point: it costs the same $2 per GPU hour as Pro and Mega, and it adds unlimited Relax mode, which queues your jobs in a slower shared queue at no extra GPU cost. Relax jobs can wait anywhere from 0 to 30 minutes, which is fine for learning and non-deadline work. Basic has no Relax fallback, so generation stops when its 200 minutes run out unless you buy more time.
For a beginner, Basic at $10 a month is plenty. Its roughly 250 standard prompts will last weeks if you are exploring casually. Move to Standard when you know you will generate regularly and want the unlimited Relax safety net. The functional reason to move to Pro is Stealth mode, which keeps your images private. That decision is covered in the privacy section below.
How do you sign up for Midjourney?
You sign up for Midjourney at midjourney.com with a Google or Discord account, then pick a plan and pay. Access is immediate after payment. Until August 2024, Discord was the only way to create an account; today you can sign up directly on the website and never open Discord at all.
- Go to midjourney.com and click Sign In. Choose Google or Discord. Google sign-in is the fastest option for new users.
- Click Subscribe and pick a plan. Start with Basic unless you already know you will generate daily.
- Pay through Stripe. Subscriptions auto-renew monthly or yearly unless you cancel.
- Open midjourney.com/imagine to generate. You can also browse the public community gallery without an account, but you cannot generate without a paid plan.
You can join the Midjourney Discord server for free and watch community generations there before committing. That is a good way to learn what prompts produce, but you cannot generate images in Discord without a paid plan either. To cancel later, open the Manage Subscription page and click Cancel Plan. Cancellation takes effect at the end of the current billing cycle, you keep any remaining fast time until then, and your generated images are not deleted.
Midjourney on the website vs Discord: which should you pick?
Pick the website. Midjourney has run primarily through its web app at midjourney.com since 2024, and Discord is now the alternative. Both produce identical results from the same model, so the choice is about workflow, not quality.
| Website (midjourney.com) | Discord | |
|---|---|---|
| How you prompt | Prompt bar at midjourney.com/imagine, with visual sliders for aspect ratio, stylize, and chaos | Type /imagine in a channel or a DM with the Midjourney Bot |
| Your past work | Gallery of past generations, searchable by keyword | Scattered through channel history and DMs |
| Editing | Canvas editor for regional edits, inpainting, and outpainting | U/V buttons and slash commands |
| Draft Mode | Yes, web only | No |
| Privacy in public | Your generations sit in your own workspace | Generations in public channels are visible to everyone in that channel |
| Best for | Focused work sessions and beginners | Watching what others make, batch workflows, experienced users |
Discord still has one real advantage: its public newbie channels are a live prompt library. Thousands of people generate images simultaneously, and reading their prompts teaches you faster than any guide. Use Discord to learn and the website to work. If you do generate in Discord, start in a #newbies channel, or message the Midjourney Bot directly so your generations stay out of the public channels. Note that direct messages only hide your images from the channel; on Basic and Standard they are still publicly visible in the Midjourney gallery. Only Pro and Mega with Stealth mode make them truly private.
How do you write your first Midjourney prompt?
You write your first Midjourney prompt as one clear sentence that names a subject, a medium, and a light or mood. On the web, type it in the prompt bar. In Discord, type /imagine, press Tab or space, and type it in the prompt field that appears.
Your first prompt can be simple. Try this one, borrowed from a working beginner guide:
/imagine a ceramic mug on a wooden table, morning light, minimalist photography
Midjourney will return four images in a 2x2 grid. A structure that works consistently once you move past the first try is subject + visual details + environment + style reference + parameters. Here is that same mug prompt, strengthened:
a ceramic mug on a rough-hewn wooden table, morning light streaming through a window, warm tones, shallow depth of field, film photography aesthetic, Sony A7III --ar 3:2 --q 2 --no text
Three rules make the biggest difference. First, put the most important element first, because Midjourney weights the beginning of a prompt more heavily than the end. Second, name the medium: "photography," "oil painting," "digital illustration," and "3D render" each pull the result in a different direction, and camera details like "85mm lens, shallow depth of field, natural light" sharpen photographic prompts. Third, do not overthink the first prompt. Generate something, see what comes back, and learn from the result.
If prompt craft is new to you generally, the same subject-first habit transfers to text tools. Our ChatGPT beginner guide teaches the same skill for words that this guide teaches for images: say what you want, in order, with the important part first.
How do you upscale in Midjourney?
You upscale in Midjourney by clicking the image you like in the four-image grid, then choosing Upscale. That gives you a single high-resolution version you can download. In Discord the same job is done by the U1 to U4 buttons under the grid: U1 is top-left, U2 is top-right, U3 is bottom-left, and U4 is bottom-right.
After you upscale, more options appear. Vary (Subtle) makes small changes to the image, while Vary (Strong) makes significantly different ones. Zoom Out extends the canvas outward, and Pan extends it in a specific direction. In Discord, the V1 to V4 buttons generate four new images similar to the one you picked, keeping the same general composition and style. The refresh button re-runs the same prompt from scratch, which is useful when none of the four images are working.
Do not upscale everything. Upscaling costs more GPU time, and once you upscale one image you cannot go back and vary the other three from that grid. Only upscale when you have a real candidate. In current versions the U button isolates the image so you can edit and download it at full quality, and HD mode renders it at 2K directly without a separate upscale step: any V8.2 image can be re-rendered at 2048 pixels from the editor, at a cost of 1.3 GPU minutes and about 12 seconds. Upscaled V8 images are typically around 1456 by 816 pixels at 16:9, which is enough for web use, social media, and most print up to roughly 5 by 7 inches. For posters or banners, run the result through an AI upscaler such as Topaz Gigapixel.
When you download a finished image, save it in the right format for where it will live. Our AVIF vs WebP comparison shows which format to pick for the web and when to keep a PNG or JPG instead.
Which Midjourney parameters should a beginner learn first?
A beginner should learn four Midjourney parameters first: --ar for aspect ratio, --s for stylize, --c for chaos, and --no for exclusions. Everything else can wait until you have a reason for it. Parameters go at the end of your prompt, after two hyphens.
| Parameter | What it does | Range | Default |
|---|---|---|---|
| --ar | Aspect ratio (width:height) | Any ratio | 1:1 |
| --s / --stylize | How much Midjourney's aesthetic overrides your prompt | 0 to 1000 | 100 |
| --c / --chaos | Variation between the four grid images | 0 to 100 | 0 |
| --no | Exclude listed elements | Text list | None |
| --sref [URL] | Apply the visual style from a reference image | Any public image URL | None |
| --sw | How strongly the style reference influences the result | 0 to 1000 | 100 |
| --draft | Draft Mode (web only): 24 images at lower GPU cost | Flag | Off |
| --v | Model version selector | 7, 8.1, 8.2, niji 7 | 8.2 |
Aspect ratio changes composition more than any other parameter. The default is 1:1 square, which is wrong for most real uses: add --ar 16:9 for landscape and video thumbnails, --ar 9:16 for vertical phone screens, --ar 4:5 for Instagram portraits, or --ar 3:2 for classic photo proportions. Stylize at 0 to 100 keeps the output close to your description; at 400 and above, Midjourney's aesthetic sense takes over and produces more striking images at the cost of prompt accuracy. Chaos at 0 keeps all four grid images visually related; raising it to 50 to 75 during early ideation gives you genuinely different directions to choose from. The --no parameter is a lifesaver for the two most common annoyances: --no text stops Midjourney adding random words to your images, and --no blurry helps when results keep coming out soft.
Two newer tools are worth knowing once the basics click. Draft Mode, added in June 2026 and available on the web only, generates 24 images at 512 pixels per prompt for 0.4 GPU minutes: six times the images at half the GPU cost of a standard four-image prompt. Use it at the start of a project to explore directions cheaply, then drop the --draft flag once you have a direction and generate at full quality. Style references let you point Midjourney at any publicly accessible image URL with --sref and control the strength with --sw between 0 and 1000.
What is Midjourney V8.2?
Midjourney V8.2 is the current default model. It became the default on July 24, 2026, replacing V8.1, which had held the default slot only since June 10, 2026. V8.2 brings bolder aesthetics, stronger personalization, and a new Edit Model that replaces Omni Reference, Character Reference, and Retexture.
| Version | Released | Status (October 2026) | What changed for users |
|---|---|---|---|
| V7 | April 3, 2025 | Selectable with --v 7 | Personalization on by default; Omni Reference introduced |
| V8.0 | March 17, 2026 | Alpha only, retired July 24, 2026 | Better prompt following, text rendering, native 2K HD mode |
| V8.1 | April 14, 2026 | Default June 10 to July 23, 2026 | Standard jobs about 4 to 5 times faster; 2K HD without upscaling |
| V8.2 | July 24, 2026 | Current default | Bolder aesthetics, stronger personalization, Edit Model replaces Omni Reference, Character Reference, and Retexture |
If your account predates the July switch and you never pinned a version, you are most likely running V8.2 by default today. To confirm, type /settings in Discord or check the model selector on midjourney.com, where the active version appears at the top of the panel. You can still select an older version with --v 7 or --v 8.1 on a single prompt, or set a persistent default in the settings panel. Use --v 7 if a workflow depends on Omni Reference, which has not been fully ported to V8.2, and --niji 7 for anime-style output.
The Edit Model is the practical difference in V8.2. After upscaling any image, the Canvas editor lets you change an image with plain written instructions, attach up to four reference images at once, and inpaint a selected area or outpaint beyond the canvas edge. Draw a freehand or rectangular selection over the part you want to change, describe the replacement, and only that area regenerates. Select a substantial portion of the image: selections that are too small tend to be ignored. Midjourney is not the only image model with a new 2026 release. Our Nano Banana explained covers Google's image model, which is the other branch of the AI-image cluster.
Can you use Midjourney images commercially?
Yes, you can use Midjourney images commercially on every paid plan. You own the images you create, even after you cancel. One threshold applies: if your company earned more than $1 million in gross revenue in the prior calendar year, Midjourney requires a Pro ($60 a month) or Mega ($120 a month) plan for commercial use. Individuals and companies below that threshold can use Basic or Standard for client projects, print sales, and merchandise.
Privacy is a separate question from ownership. On Basic and Standard, every image you generate is publicly visible in the community gallery by default. Stealth mode, exclusive to Pro and Mega, makes your generated images private instead. If privacy matters to your workflow, Stealth mode is the primary functional reason to upgrade from Standard to Pro. One more ownership edge case: images you upscale from another member's work belong to that member, not to you.
7 beginner mistakes that waste your fast GPU time
Most wasted GPU time comes from seven repeatable mistakes. Avoid them and a Basic plan lasts far longer.
- Prompt too vague. "A cool logo" tells Midjourney almost nothing. The more specific your description, the more useful the output.
- Giving up after the first four images. The first generation is rarely the final result. Click V on a close-but-not-quite image and keep iterating.
- Not trying different aspect ratios. The default 1:1 square crops out a lot of interesting compositions. Try --ar 16:9 or --ar 4:5 and see how the same prompt feels completely different.
- Prompting for specific text in the image. Midjourney is still rough at rendering actual words. Generate without text and add it in Canva, Photoshop, or Figma afterward, or put the exact text in double quotes and accept that it may need a retry.
- Upscaling every single image. Upscaling costs more GPU time and you cannot go back to vary the others. Only upscale when you have found a real candidate.
- Generating in public Discord channels by default. Your images are visible to everyone in the channel, and on Basic and Standard they are public in the gallery too. Use the website or a bot DM for anything you would rather keep quiet.
- Expecting photorealistic product photos. Midjourney does stylized imagery incredibly well. For straight product photography where you need accuracy and no hallucinated details, Adobe Firefly or a traditional photographer will serve you better.
Your first week with Midjourney
One week of short sessions is enough to go from first prompt to a repeatable workflow. This plan spends about 30 GPU minutes total on Basic, leaving the rest of the month untouched.
| Day | Do this | What you learn |
|---|---|---|
| Day 1 | Sign up, subscribe to Basic, run 5 simple prompts on the web | The prompt bar, the 4-image grid, download |
| Day 2 | Re-run your best prompt with --ar 16:9 and --ar 4:5 | How aspect ratio changes composition |
| Day 3 | Pick one image, upscale it, then Vary (Subtle) twice | Upscale vs vary, and why you only upscale candidates |
| Day 4 | Try Draft Mode with --draft on one idea, 24 images at once | Cheap direction-finding before full-quality generation |
| Day 5 | Open your best image in the Canvas editor and change one region | Regional editing and the Edit Model |
| Day 6 | Re-run a prompt with --s 50, then --s 400, and compare | How stylize trades accuracy for beauty |
| Day 7 | Generate one finished image for a real use: a thumbnail, a moodboard, or a social post | The full loop, end to end |
When the week is done, you will know whether Midjourney earns a permanent place in your workflow or whether a different tool fits better. If your next project is video rather than still images, our Google Vids beginner guide covers the AI-video branch of the same creator workflow.
Sources
- AI Weekly, "How to Use Midjourney in 2026: Plans, Prompts, and V8.1" (aiweekly.co) — plans, GPU-minute costs, V8.1 default date, Draft Mode, parameters; last verified July 25, 2026 against official docs and pricing pages.
- Albato, "Midjourney Pricing 2026 + How to Sign Up (All Plans)" (albato.com) — plan table, V8.2 default July 24, 2026, Edit Model, concurrent jobs, cancellation terms, niji journey trial note; prices sourced from Midjourney's official plan comparison.
- TechSifted, "How to Use Midjourney: Step-by-Step Beginner's Guide (2026)" (dev.to) — account setup, first prompt, U/V buttons, common mistakes, commercial-use threshold; updated September 5, 2026.
- Midjourney version documentation, via the Albato and AI Weekly citations above — V8.2 default July 24, 2026; V8.1 default June 10 to July 23, 2026.
Frequently asked questions
Is Midjourney free?
No. Midjourney has no free plan and no free trial on its website or in Discord. The free trial ended in 2023. The cheapest way in is the Basic plan at $10 a month, and the only trial anywhere is a limited one inside the separate niji journey mobile app. You can browse the public gallery and watch Discord generations for free, but generating images requires a paid subscription.
Do you need Discord to use Midjourney?
No. Since the web interface opened to everyone in August 2024, midjourney.com/imagine is the primary way to use Midjourney. Discord is still available as an alternative and is useful for watching what other people make, but you can sign up with a Google account and generate entirely in the browser.
Which Midjourney plan should a beginner pick?
Basic at $10 a month. It includes 3.3 hours of fast GPU time, about 250 standard image prompts a month, which lasts weeks of casual learning. Move to Standard at $30 a month when you generate regularly: it adds unlimited Relax mode and the same $2-per-GPU-hour rate as the higher plans. Move to Pro only when you need Stealth mode to keep your images private.
What is the default Midjourney version in 2026?
V8.2, which became the default on July 24, 2026. V8.1 was the default only from June 10 to July 23, 2026. V8.2 adds bolder aesthetics, stronger personalization, and an Edit Model that replaces Omni Reference, Character Reference, and Retexture. You can still select V7 or V8.1 with --v 7 or --v 8.1 if an older workflow depends on them.
Can I use Midjourney images commercially?
Yes, on every paid plan. You own the images you create, even after you cancel. The one exception is companies that earned more than $1 million in gross revenue in the prior calendar year: they need a Pro or Mega plan for commercial use. If your images must stay private rather than visible in the public gallery, you also need Pro or Mega for Stealth mode.