how to use midjourney
Use Midjourney V8.2 from first prompt to refined image, with the right controls, seven workflows, honest limits, and three business opportunities.
You can turn a short description into four visual directions, refine the strongest one, carry a chosen look into the next batch, and even animate a still without leaving Midjourney's web workspace. The current default is V8.2, released on July 24, 2026, so the useful way to learn Midjourney now is as a repeatable creative workflow, not a bag of Discord commands.
About 720 US Google searches a month ask exactly how to use Midjourney. The short answer is: subscribe, open the Create page, describe the image you want, choose from four results, then refine the winner. The better answer is to decide what each input is responsible for before you generate anything. Text describes the subject. An Image Prompt guides content and composition. A Style Reference carries the visual treatment. Personalization carries your taste across jobs.
That division of labor is what turns random prompting into useful creative direction.
What Midjourney actually does
Midjourney is best treated as a visual exploration engine. You give it a prompt, which is simply the text describing what you want to see, and it returns a set of four images. You can then make subtle or strong variations, change the prompt with Remix, enlarge a keeper, edit a region, expand the canvas, or turn one image into an initial five-second video.
The analogy is a fast concept room with four art directors sketching at once. They are good at opening directions you may not have considered. They are not a production department that guarantees the same bottle label, character face, floor plan, or legal clearance every time.
Midjourney V8.2 is the current default image model. Midjourney describes the update as focused on aesthetics, image quality, and Personalization, its system for learning what you tend to like. V8.2 also supports 2048-pixel HD image generation. Standard jobs cost about 0.8 GPU minutes per four-image set, while HD jobs cost about 1.3. GPU time is the metered processing allowance included with a subscription, much like data in a mobile plan.
How to use Midjourney, step by step
The cleanest starting path is the website. Discord still works, but a beginner does not need to learn chat commands before learning visual direction.
1. Pick a plan and open Create
Log in at midjourney.com, choose a subscription, and open the Create page. The regular workflow is paid. Basic starts at $10 a month; Standard, Pro, and Mega cost $30, $60, and $120. If plan limits matter, the full Midjourney pricing breakdown explains Fast time, Relax mode, and Stealth without turning this guide into a billing page.
2. Set the canvas before the prompt
Open the settings beside the Imagine bar. Choose the aspect ratio for the place the image will live, such as square for a product tile, portrait for a cover, or landscape for a presentation. Confirm V8.2, then choose SD for exploration or HD when the extra resolution is worth the higher GPU cost. V8.2 allows aspect ratios up to 14:1 in SD and 4:1 in HD.
3. Write the prompt as a shot brief
Start with the subject, then add medium, setting, lighting, mood, and composition only where they matter. Midjourney itself recommends short, precise prompts over long instruction lists. A useful pattern is: subject + medium + environment + light + composition + mood.
For example: "ceramic desk lamp, editorial product photograph, warm stone studio, soft side light, three-quarter view, quiet and tactile." This tells the model what belongs in the frame. It does not bury the subject under a paragraph of wishes.
4. Generate four options and diagnose the miss
Press Enter or Send. Midjourney returns four images. Do not rewrite everything when the batch misses. Name the failing dimension. If the subject is wrong, clarify the subject. If the framing is wrong, change the composition. If the look is wrong, use a Style Reference. If one image is close, make a variation instead of starting over.
5. Refine the keeper
Use Variations for alternate takes, Remix to change the prompt while preserving the direction, and Subtle or Creative Upscale when you need a larger image. The Editor combines canvas expansion, zooming out, and region edits. One technical catch matters: editing an HD V8.2 image downscales that edited result to SD, after which you can upscale again.
6. Save the repeatable parts
Pin reusable references in the Imagine bar, create folders for projects, and turn on Personalization only when you want your learned taste to affect the job. For a specific campaign look, a Style Reference is usually more controllable than a global preference profile.
The four controls people confuse
Each reference tool answers a different question. Mixing them indiscriminately is the fastest route to muddy results.
An Image Prompt is inspiration, not a photocopier. Midjourney supports one image plus text, multiple images without text, or multiple images plus text. A single image cannot stand alone as the whole prompt. A Style Reference is narrower: it carries the look, but it does not copy the reference's people or objects. Personalization learns from the images you select and like, then applies that taste more broadly.
Seven Midjourney use cases, ranked by commercial value
The highest-value uses happen before expensive production, when seeing several credible directions changes a decision.
1. Ecommerce campaign previsualization
An ecommerce creative lead can upload a product photo as an Image Prompt, add a campaign brief, and explore several environments before booking a photographer. A Style Reference can keep the lighting and palette aligned across concepts. The payoff is a sharper shot list and fewer vague conversations with a production team.
Use this for concepts, not final pack shots. Midjourney says Image Prompts inspire new images rather than copy the source exactly, so packaging, labels, proportions, and small product details still need human verification.
2. Agency pitch concepts
An ad agency pitching three campaign territories can give each territory its own Style Reference, generate key visuals, and use variations to test tone without commissioning three finished campaigns. The client reacts to something concrete, while the agency learns which direction deserves production money.
The payoff is decision speed. The deliverable is still a concept board, not evidence that the final shoot, animation, or typography is solved.
3. Book-cover art direction
An indie author or small publisher can translate a manuscript's genre, setting, emotional temperature, and central symbol into several cover-art directions. The strongest image then moves to a designer for title treatment, hierarchy, spine, back cover, and print specifications.
This works because Midjourney is being asked for art direction, not a finished publishing file. It reduces the expensive part of early exploration while keeping typography and market judgment in human hands.
4. Storyboards and previsualization
A director, producer, or UGC ad team can split a script into beats, prompt one frame per beat, and use a shared Style Reference to keep the visual language related. A selected frame can become a five-second motion test, and videos can be extended in four-second increments up to 21 seconds.
The payoff is alignment before filming. Camera, wardrobe, prop, and location conversations become specific. Character continuity is still a constraint: V8.2 does not use Omni Reference natively, and selecting it moves the job to V7.
5. Interior concept boards
An interior designer can explore material combinations, lighting moods, and furniture language for a room before sourcing. An Image Prompt can carry a rough composition, while a Style Reference can hold the material and photographic treatment.
The payoff is faster preference discovery with a client. The output is not a measured plan, a construction drawing, or proof that the shown furniture exists.
6. Game and entertainment world exploration
An art director can generate environments, props, factions, and color scripts around one creative brief, then organize keepers by project folder. Variations help explore nearby options without losing the central direction.
The payoff is breadth at the earliest stage. The catch is consistency across dozens of production assets, where a dedicated character or asset pipeline still matters more than a beautiful single frame.
7. Social campaign variations
A small content team can create a hero still, use variations to find alternate crops and arrangements, and animate a selected image for a short loop. Choosing the target aspect ratio before generation keeps the composition native to the placement instead of forcing a later crop.
The payoff is a family of related campaign concepts from one direction. The team still needs brand review, factual review, and accessibility work before publishing.
If your real question is whether Midjourney is the right engine at all, the 2026 AI image-generator comparison puts it beside the alternatives.
Three things worth building around this workflow
Midjourney is a poor foundation for an automated wrapper because its current terms prohibit automated tools from accessing, interacting with, or generating through the service. The viable opportunities are human-operated services, internal creative systems, and finished deliverables where Midjourney handles exploration rather than the whole product.
Strongest opportunity: an indie-author cover concept studio
Build a productized service that turns a manuscript brief into three market-aware cover directions, then finishes the selected direction with human typography and production layout. The buyer is an indie author or small press that needs more than a generic template but is not ready to manage an open-ended design engagement.
The demand is unusually clean: "ai book cover generator" gets about 1,600 US Google searches a month, with keyword difficulty 11 and transactional intent. "Canva AI book cover generator" adds another 880 monthly searches. Book Brush already charges $149 a year for author-focused cover tools, which proves buyers pay for a structured workflow around this job.
The smallest sellable version is an intake form for genre, synopsis, comparable covers, and non-negotiable symbols; three Midjourney art directions made by a human operator; one revision round; then a designer-built ebook cover. Add print wrap only after the core offer works.
The catch is also the moat: final typography, genre judgment, source-image rights, and disclosure expectations cannot be delegated to a pretty image. A pure prompt service is easy to copy. A trusted cover process with human finish is not.
A product campaign previsualization desk
Build a concept service for Shopify brands that turns one product photo and a brand board into six campaign territories, each with a hero concept, supporting compositions, and a practical shot list. The buyer is a founder or creative lead planning a real shoot.
"Ai product photo generator" gets about 90 US Google searches a month at a $10.23 CPC, and "best ai product photo generator" gets another 50. Pebblely charges $9 to $39 a month for automated product-photo generation, so the commodity layer is already priced. The better offer is the thinking around the image: what to shoot, which setting fits the brand, and which concept deserves production.
The MVP is a fixed brief, six concepts, and a one-page production recommendation. The honest catch is product fidelity. Midjourney treats uploaded images as inspiration, so do not promise label-perfect final ads.
A script-to-storyboard service for short ads
Build a human-operated previsualization service that turns a 30 to 60-second script into a beat sheet, 8 to 12 frames, and two short motion tests. The buyer is a small agency, creator team, or founder who needs stakeholder approval before paying for production.
"Ai storyboard generator from script" gets about 50 US Google searches a month, with difficulty 12 and transactional intent. Katalist prices its storyboarding product from $29 a month when billed annually. That is modest demand, but the buyer is close to a production decision and the output is easy to evaluate.
The MVP needs a script intake, a visual reference board, one consistent frame template, and a revision pass. The catch is continuity across shots. It also must stay human-operated under Midjourney's automation rule. Sell a directed storyboard, not unlimited generations.
What Midjourney does not solve
Midjourney does not remove the need for taste, verification, production craft, or rights review. It makes exploration cheap enough that you can compare more directions before committing.
The hard limits are practical:
- It is not an automation backend. The terms prohibit automated access and automated generation, competitive products built through the service, and resale of service access.
- References are not identity locks. Image Prompts guide rather than copy. Style References carry treatment rather than people or objects. V8.2 sends Omni Reference work to V7.
- Editing can change the resolution path. Editing an HD V8.2 image returns an SD result that needs another upscale.
- Privacy costs more. Content is open and remixable by default in public settings. Stealth is limited to Pro and Mega, and anything made in a shared Discord space remains visible in that space.
- Ownership is conditional, not magical. Midjourney's terms generally give subscribers ownership to the extent allowed by law, subject to third-party rights. Companies above $1 million in annual revenue need Pro or Mega. Midjourney also receives a broad license to submitted content and generated assets.
- The service can change. Midjourney does not guarantee stability, uptime, reliability, or that a particular model behavior will stay the same.
Use Midjourney when the value is in discovering and communicating a visual direction. Do not use it as the final authority for brand-critical text, exact products, regulated claims, construction detail, or legal clearance.
Frequently asked questions
How can I use Midjourney?
Subscribe at midjourney.com, open the Create page, type a visual brief into the Imagine bar, and submit it. Midjourney returns four images. Choose the closest direction, then use variations, Remix, upscale, or the Editor to refine it.
Can I use Midjourney for free?
The regular web and Discord workflow requires a subscription. The four monthly plans start at $10 for Basic and rise to $120 for Mega.
How to use Midjourney for best results?
Write a compact shot brief: subject, medium, environment, lighting, composition, and mood. Add only the details that matter. Use an Image Prompt for content guidance, a Style Reference for visual treatment, and variations when one result is already close.
Can I sell images I create with Midjourney?
Midjourney's terms generally give paid subscribers ownership of their generated assets to the extent allowed by law, subject to the terms and third-party rights. A company above $1 million in annual revenue must be on Pro or Mega. Ownership of a file does not guarantee copyright protection or clear third-party rights, so commercial work still needs review.
Does Midjourney own my images?
The terms generally say you own assets you create to the extent allowed by law, with stated exceptions. They also grant Midjourney a broad, continuing license to content you submit and assets the service produces. Treat privacy and usage rights as separate questions.
If you want one of these human-operated creative systems built for your business, see the AI services.
Aug 5, 2026
