How to Use Midjourney (2026): Prompts, Plans, Rights

Midjourney is still the tool people mean when they say “AI images that look designed.” It is also the tool that traps newcomers with two myths: that it is free, and that Discord is mandatory. Neither is true in 2026.

This is a practical how-to: account and plans, web vs Discord, first good generations, prompt structure, iteration habits, commercial rights questions, and when to use something else. Product verdict and pricing shapes: Midjourney review. Ranked category: best AI image generators and the image hub. Craft: how to write AI image prompts. Head-to-head with the model many people confuse it with: Midjourney vs FLUX.

Plan names, Fast minutes, and Terms of Service change. Figures below match our review notes (pricing checked 2026-07-27). Confirm on midjourney.com before you pay.

What Midjourney is (and is not)

Is: a paid image generation product with a strong house aesthetic, excellent for concept art, mood boards, editorial vibes, and stylised marketing exploration. Web app + Discord access. Fast iteration once you learn references and parameters.

Is not: a free Canva substitute; a reliable typography engine; a model API you self-host; a guarantee of commercial safety equal to Adobe’s Firefly licensing story; a literal layout engine that obeys pixel-perfect wireframes.

If you need readable headlines in the frame, start at Ideogram (and read the image generators guide). If you need API / open weights / photoreal + text as a model, read FLUX and how to use Flux AI. If you need finished social templates, Canva may beat raw Midjourney.

Honesty first: there is no free tier

Free trials ended years ago. The real entry price is a paid subscription. Our review last checked:

Exact Fast minutes, GPU time, and concurrent jobs belong on Midjourney’s live pricing page — treat the table as shape, not a contract.

Implications:

• You cannot “just try one image free” inside Midjourney itself the way Ideogram or Firefly free credits allow.
• Budget learning generations as part of month one; your first fifty images are tuition.
• If $10 is too much to learn, practice prompt craft on free-credit tools, then move to Midjourney when aesthetics matter enough to pay.

Web app vs Discord

Web app (default recommendation for new users)

1. Create an account on midjourney.com and subscribe.
2. Open the create / imagine interface in the browser.
3. Enter a prompt, generate, rate, vary, upscale (wording of controls evolves — look for create, vary, remix, organize).
4. Use the gallery/archive to build a personal library of winners and their prompts.

Why start here: no server etiquette, no bot commands to memorise, clearer personal archive for most people.

Discord (optional)

Discord remains available for users who like channel energy, shared rooms, or old muscle memory (/imagine-style flows). It is not required to get value in 2026. If public channels stress you, use web + privacy features on higher plans.

Practical tip: pick one surface for a week. Splitting attention between Discord habits and web UI slows learning more than either UI alone.

Your first hour (workflow that does not waste the month)

1. Pick a real job — “hero image for a landing page about quiet productivity apps,” not “cool fantasy.”
2. Write a five-slot prompt (below) — one paragraph, not twenty adjectives.
3. Generate a grid, pick the closest mood, not the perfect subject.
4. Vary / remix that image; change one slot (light, or lens, or background).
5. Save the prompt + seed/reference habits Midjourney exposes in your UI — the library is the asset.
6. Stop at a good-enough direction and only then chase detail. Endless re-rolls without a locked direction burn Fast time.

This is the same discipline as our general image prompt guide: specificity over keyword spam.

Prompt structure that works in Midjourney

Midjourney rewards photographic or artistic specificity. It punishes synonym piles (“cinematic epic moody 8k masterpiece”) that pull toward average.

Example prompt:

tired barista in a linen apron wiping the bar between rushes, narrow coffee bar with rainy street window, medium shot, 35mm, eye level, cool window light mixed with warm interior practicals, photoreal, natural skin texture

Then iterate: same prompt, swap only light; or only lens; or only “illustration” vs “photoreal.”

Midjourney-flavoured levers (concepts, not a parameter manual)

UI labels change; the ideas stay:

• Style reference / image reference — lock a look you already like instead of re-describing it every time.
• Aesthetic strength — how hard the house style pushes; dial when the look fights a literal brief.
• Aspect ratio — decide deliverable shape first (1:1 social, 16:9 hero, 9:16 story).
• Stylize vs raw-ish control — beauty vs obedience trade-off; know which you are optimising.
• Vary region / inpaint-style edits when available — fix one object without rerolling the universe.

For a parameter cheat you will re-check on official docs: Midjourney’s own documentation on the site and help centre is authoritative; third-party “parameter lists” rot quickly.

Iteration habits of people who get consistent output

Do:

• Keep a prompt library with notes (“too warm,” “hands failed,” “logo unreadable — expected”).
• Use references from your own best outputs for campaign consistency.
• Generate mood first, layout second.
• Separate exploration Fast from final Fast so you do not burn the month on experiments in highest quality by accident.

Do not:

• Change subject, style, and lighting in one hop — you learn nothing.
• Chase perfect hands/text for hours when Ideogram or a human designer is cheaper.
• Paste client confidential brand packs into public Discord contexts.
• Assume upscale fixes a bad composition — it sharpens the mistake.

Text, posters, and logos (set expectations)

Midjourney remains weaker at readable text than Ideogram and often weaker than FLUX-class models. For:

• Posters with headlines → Ideogram (or design tool + type)
• Logo exploration → Ideogram for letterforms, then a human in vectors; never ship a raster “logo” as final mark without design process
• Pure mood / key art without words → Midjourney shines

Hybrid workflow: Midjourney for atmosphere stills → designer adds type in Figma/Photoshop → or Firefly generative fill inside Adobe for licensed-stack edits (Firefly, design guide).

Plans: how to choose without overbuying

Start Basic if you have never used Midjourney and only need to validate whether the aesthetic fits your brand.

Move to Standard when Relax unlimited-style generation matters and Fast alone is too tight for your weekly volume.

Pro when privacy (Stealth-class features) or Fast ceiling is the pain — not because Pro sounds senior.

Annual vs monthly: if you already know you will stay, annual discounts often exist; do not annualise a tool you have not stress-tested for two weeks.

Team / org: check Midjourney’s current team offerings and who owns the generations under company policy. Personal Pro on a work brand can become an ownership mess.

Commercial rights and client work (read the live terms)

PromptHive is not a law firm. What we can say as operators:

1. Licence follows plan and Terms — read Midjourney’s Terms of Service / docs for your plan.
2. Paid ≠ “use anywhere forever without thought.” Trademarked characters, celebrity likenesses, and client brand misuse are still your problem.
3. Adobe Firefly markets a different commercial story (licensed training data positioning for Firefly models). If legal needs that story, compare Firefly rather than forcing Midjourney to be Firefly. Partner models inside Firefly may not share the same promise — check which model ran.
4. Agency rule of thumb: get written client acceptance of AI tooling and keep generation metadata/prompts for the file trail.
5. Stock and “AI stock” marketplaces may have their own rules for Midjourney-origin art — read before you sell.

When in doubt for a high-value campaign, commission a human illustrator or use the stack legal already approved.

Midjourney vs the rest (when to leave the app)

Category decision tree: best AI image generators.

Common failures and fixes

Ethics and disclosure notes

Platform and regional rules on AI content labelling evolve (especially for video and realistic media — see complete AI video guide for the stricter end of the spectrum). For still marketing images, still check:

• Client and platform policies
• Whether synthetic imagery must be disclosed in your market
• Whether a realistic person-like image could be mistaken for a real endorsement

Do not generate non-consensual deepfake-like portraits of real private individuals for publication.

Building a repeatable brand look in Midjourney

One beautiful image is easy. A campaign set that looks like one brand is the actual job.

1. Write five non-negotiables before any prompt: colour temperature, lens family, how much realism vs illustration, what never appears (no neon, no stock handshake, etc.), and aspect ratios you actually ship.
2. Generate twenty exploration images on Basic Fast or Standard Relax — quantity is tuition.
3. Pick one hero as the style or image reference for the next twenty.
4. Vary only subject or only setting inside that lock.
5. Export a board for stakeholders; freeze the reference before anyone “just tries a totally different vibe.”
6. Add type and logo in a design tool — Figma, Photoshop, Canva — rather than asking Midjourney to spell the brand.

If stakeholders cannot agree on the five non-negotiables, no parameter will invent taste. Design systems still matter; Midjourney is not a replacement for one. Wider craft: complete AI design guide.

Midjourney for specific jobs

Mood boards and concept art

This is the home turf. Prompt for emotion and material, not for final layout grids. Deliver a range, not a single “final.” Human art direction still picks the winner.

Website and product heroes

Generate wide ratios early. Expect to crop. Keep negative space if you will overlay UI or headlines later — models love filling every corner with detail you will paint out.

Social thumbnails

Aspect ratio first. If the thumbnail needs words, generate the scene in Midjourney and set type in Canva/Figma, or switch the whole job to Ideogram. Fighting Midjourney for “SALE 50%” is how Fast time dies.

Print and packaging exploration

Useful for mood and material studies. Not a substitute for dielines, print profiles, or trademark clearance. Hand off to a designer before production files.

Character consistency

Better than early generations, still imperfect across many poses and scenes. For narrative series, plan on references, limited wardrobe changes, and human cleanup. Video multi-shot consistency is a harder problem in another category (video guide).

Organising your library (underrated)

Treat the archive like a design system:

• Folders or albums by campaign, not by date only
• Prompt text stored next to the file (spreadsheet or note)
• Tags for “usable final,” “reference only,” “failed hands,” “too on-model Midjourney”
• A short “do not repeat” list of prompts that always waste Fast

When a client asks “can we get more like image 14?” you need to find image 14 and its prompt in seconds. The people who “can never recreate that look” usually never saved it.

Fast vs Relax (how to think about speed)

Exact names and allowances change; the economic idea is stable:

• Fast spends the scarce monthly resource for snappy iteration
• Relax (on plans that include unlimited-style Relax) trades waiting for not burning Fast

Healthy pattern: explore composition in Relax when available; spend Fast when you are refining a locked direction under a deadline. Unhealthy pattern: every random experiment in Fast because waiting feels annoying — that is how Basic feels empty on day twelve.

Team norms if more than one person has access

• Decide who owns the subscription and who owns the files
• Ban pasting confidential product shots into shared public surfaces
• Agree whether Stealth/privacy features are required for unreleased work
• Share the prompt library, not only the JPEGs
• Review commercial terms once per quarter when client work is involved

Personal accounts used for company brands create ownership and offboarding messes. Sort that before the rebrand ships.

Learning path for absolute beginners

Day 1: Subscribe, web app only, five-slot prompts, ten images, no parameters obsession.
Day 2: Pick one winner; vary only light; vary only lens.
Day 3: Learn aspect ratios for your real channels.
Day 4: Read prompt craft and delete adjective spam from your defaults.
Day 5: Try one Ideogram text job so you know when to leave Midjourney.
Day 6: Skim Midjourney vs FLUX so you stop comparing a product to a model unfairly.
Day 7: Build a twelve-image board for a real project; stop when the board works, not when the meter hits zero.

The short version

1. Pay to play — no free Midjourney tier; Basic ~$10 class is the real door.
2. Use the web app unless you prefer Discord.
3. Prompt with five slots; stop stacking empty adjectives.
4. Iterate one variable at a time; save winners.
5. Do not use Midjourney for critical text-in-image.
6. Read commercial terms for your plan before client delivery.
7. Compare FLUX, image hub, and prompt craft when Midjourney is the wrong job.

Where to go next

• Midjourney review — pros, cons, pricing shape
• Midjourney vs FLUX
• How to write AI image prompts
• Best AI image generators · Image hub
• How to use Flux AI
• Complete AI design guide
• Ideogram · Firefly · Canva