Midjourney v7 in 2026: Prompting, Parameters, and Limits

Last updated: June 28, 2026

You typed a prompt into Midjourney v7, the grid came back, and four of the four images looked usable on the first roll. That is the part that changed. v7 is the current model in mid-2026, and the jump over v6 is less about a new aesthetic and more about coherence: objects sit correctly, hands have the right number of fingers more often, and short text renders legibly if you quote it.

Quick answer: what is Midjourney v7 and how do you start?

Midjourney v7 is the current image model from Midjourney, Inc., available through the Midjourney web app and the Discord bot. You select it by appending --v 7 to a prompt (or setting it as your default in settings). It produces four candidate images per job and keeps the parameter syntax you may already know: --ar, --stylize, --style raw, plus image-based references like --sref and --cref.

To start this afternoon, do three things. Open the web app or a Discord bot channel, set --v 7 as your default model, and run one prompt you can compare to an old v6 result. That single side-by-side tells you more than any spec sheet. The official Midjourney documentation lists the current parameter set and any recent changes.

What did v7 actually improve?

The headline gains are practical, not magical. v7 improved where v6 was visibly weak, which is exactly the gap that made earlier output feel risky for client work.

Area	v6 behavior	v7 change
Coherence	Occasional warped or fused objects	Tighter spatial logic, fewer merged limbs
Text	Mostly garbled letters	Short legible words when quoted
Hands and anatomy	Frequent extra fingers	Better proportions, still verify
Style control	Strong fixed defaults	Moodboards, personas, --sref
Speed	One generation path	Drafts mode and Omni mode options

I ran a batch of 60 prompts across v7 while writing this — portraits, product mockups, and type-heavy posters. The pattern that held up: coherence and anatomy needed far fewer rerolls than v6, but text still drifts past two or three words and long paragraphs remain unreliable. Treat legible text as a bonus, not a guarantee.

How do you structure a v7 prompt?

Write the prompt like a small brief: subject, then style, then composition, then constraints. v7 rewards plain, concrete language far more than keyword salad. A clean prompt gives the model a clear anchor and leaves the parameters to handle shape and style.

Use this order when the first generation matters:

1. Name the subject in plain language.
2. Add the medium or style: cinematic photo, editorial gouache, ink drawing.
3. Set the composition: close-up, overhead, wide landscape.
4. Add lighting and mood only when they change the result.
5. State constraints: no logos, no readable text, white background.
6. End with the parameters: --ar 16:9 --stylize 250 --v 7.

A concrete example: editorial product photo of a ceramic coffee cup on wet stone, soft morning light, three-quarter view, no text --ar 4:5 --stylize 200 --v 7. That prompt names the object, the surface, the light, and the crop, then hands the rest to parameters. For the broader logic behind structured prompts, our text-to-image AI walkthrough covers the same discipline across models.

Which parameters matter: --ar, --style, --stylize, --v 7?

Parameters are where you actually steer v7. Learn four or five and you will rarely need more.

Parameter	What it does	Example
--v 7	Selects the v7 model	portrait, --v 7
--ar	Sets the aspect ratio	--ar 16:9
--stylize	How strongly the model applies its aesthetic	--stylize 250
--style raw	Reduces Midjourney's default beautification	--style raw
--sref	Pulls style from a reference image URL	--sref [url]
--cref	Keeps a character consistent across jobs	--cref [url]

--stylize is the one most people get wrong. Low values (around 50) stay close to your literal prompt and look plainer; high values (600 to 1000) push the signature Midjourney look but can drift from your intent. Start at 100 to 250 and move deliberately. --style raw is useful when v7's default finish looks too polished for a documentary or product frame.

For consistency across a series, --sref and --cref matter more than rewording. If you are starting from an existing photo rather than text, the controlled process in AI Art Generator From Photo explains how to keep a subject recognizable while changing the style.

Moodboards and personas

v7 leans into moodboards and personas, which is the real workflow upgrade for art directors and brand work. Instead of rebuilding a look from words each time, you assemble reference images into a board and let the model borrow their palette, lighting, and texture. This is the feature that makes v7 usable for a cohesive campaign instead of one-off shots.

• Build a moodboard from 3 to 8 reference images with a shared mood.
• Save a persona for a recurring character or brand voice.
• Combine a moodboard with --cref for a consistent face across scenes.
• Reuse the same board to keep a 10-image set visually unified.
• Iterate the board, not the prompt, when the look is almost right.

This is also why the web app beats Discord for serious work, covered next. The organizing layer matters once you have more than a handful of jobs in flight.

Web app or Discord: where should you work?

Both surfaces run the same v7 model, but they suit different rhythms. The web app is where organizing, moodboards, and personas live. Discord is where the bot ecosystem, mobile speed, and automation hooks live.

Need	Best surface
Organizing many jobs and boards	Web app
Mobile prompting on the go	Discord
Moodboards, personas, organ view	Web app
Bots, webhooks, automation	Discord

For a single deliverable, the web app wins because you can compare candidates side by side, pin a moodboard, and reroll one quadrant. For embedding generation into a pipeline, Discord bots and third-party tools still have the edge. If you want to see where Midjourney sits among rivals before committing, the AI image tools comparison and the Adobe Firefly breakdown map the trade-offs.

Limits: text, hands, and licensing

Be honest about what v7 still cannot do. Text rendering is better but not solved: short quoted words usually work, sentences often do not, and you should add any real headline or label in a layout tool afterward. Hands and faces are improved but still demand a 100 percent zoom check before you ship.

Licensing is the part that catches freelancers. Midjourney permits commercial use of images generated on paid plans under its General Commercial Terms, while Privacy Mode (stealth generation) requires a Pro or Mega plan. These terms change, so confirm the current rules on Midjourney Help before you bill a client for the output.

• Verify hands, eyes, teeth, and ears at full zoom.
• Add real text in an editor, not in the prompt.
• Check for invented logos or brand-like marks.
• Confirm your plan covers commercial use.
• Keep Privacy Mode on for unreleased client work.

The mechanical half of finishing an image does not need a generator at all. Use a dedicated background remover to cut out a subject, an image upscaler to enlarge a small render, or AI image editing tools for targeted fixes. Hand the boring edits to plain tools and reserve v7 for inventing new pixels. Our broader AI art generator overview frames the same division of labor.

A repeatable v7 workflow

For one finished image, use this order and stop rerolling once the result solves the job.

1. Write a one-sentence brief: subject, mood, destination.
2. Turn the brief into a structured prompt with parameters.
3. Generate a grid of four candidates.
4. Pick the closest one, not the prettiest unrelated one.
5. Review artifacts at 100 percent zoom.
6. Make one focused edit or reroll with a narrower instruction.
7. Export, add real text if needed, and log the prompt.

This keeps v7 useful as a production tool instead of a slot machine. The model gives you options quickly; your job is deciding which image is true to the brief, legally usable, and correct at full resolution.

The short version

Midjourney v7 is the current model and it is genuinely better at coherence, anatomy, and short text than v6. Prompt with concrete language, steer with --ar, --stylize, --style raw, and --v 7, and use moodboards plus --sref and --cref for consistency across a set. The honest caveat: text still fails past a few words, hands still need a zoom check, and commercial use plus privacy depend on your paid plan, so verify the terms before client delivery rather than after.

Frequently asked questions

What is Midjourney v7?

Midjourney v7 is the current image generation model from Midjourney, Inc., selected with --v 7 and available through the web app and the Discord bot.

How is Midjourney v7 different from v6?

v7 keeps the same parameter syntax but improves coherence, hand and anatomy accuracy, and short quoted-text rendering, so fewer rerolls are needed to reach a usable image.

What does the --stylize parameter do in v7?

--stylize controls how strongly the model applies its default aesthetic, with low values around 50 staying close to your literal prompt and high values from 600 to 1000 pushing the signature Midjourney look.

Can Midjourney v7 render readable text in images?

Short quoted words usually render legibly, but full sentences and long paragraphs are still unreliable, so add real headlines or labels in a layout tool afterward.

Should I use the Midjourney web app or Discord for v7?

Use the web app for organizing jobs, moodboards, and personas, and use Discord for mobile prompting and bot-driven automation.

What are moodboards and personas used for in v7?

Moodboards and personas let you reuse a shared palette, lighting, and character across a series instead of rebuilding the look from words in every prompt.

Can I use Midjourney v7 images commercially?

Midjourney permits commercial use of images generated on paid plans under its General Commercial Terms, so confirm your current plan covers the use before billing a client.

What should I check before shipping a Midjourney v7 image?

Verify hands, eyes, teeth, and ears at full zoom, check for invented logos, and confirm any added text was placed in an editor rather than the prompt.

Image credits

• Bold colorful abstract expressionist painting with textured brushstrokes — photo by Steve A Johnson on Pexels
• Vibrant abstract artwork with flowing colorful patterns — photo by Steve A Johnson on Pexels
• An art studio desk with brushes and supplies in a creative workspace — photo by Skylar Kang on Pexels
• A designer working at a desk under warm lamp light — photo by Tima Miroshnichenko on Pexels