Loading tutorials…
Loading tutorials…
Your prompt looks fine. The output is "okay." But it does not feel like YOUR brand — it feels like generic AI imagery. This walks through the systematic diagnosis specialists run.
Who this is forAnyone past the Midjourney setup phase whose output keeps getting rejected as "generic" or "AI-looking." Usually the fix is one of 6 specific issues. This tutorial finds yours.
What you'll need
Step 1
Wrong stylize is the #1 cause of generic output. Too high = artistic mush. Too low = stiff stock-photo feel. Match to use case.
In your prompt, find the --stylize value. If not specified, it defaults to 100.
--stylize 100 for ad creative is usually fine. For product photography it is too artistic; for brand-mood imagery it is too literal.
Diagnose by use case: Product/e-commerce → try --stylize 25-50. Editorial/lifestyle → try --stylize 100. Brand mood/hero → try --stylize 250-500. Illustrated/conceptual → try --stylize 500-1000.
Re-generate the same prompt with different stylize values to see the spread. The right value will be obvious.
Step 2
Without --sref, every generation is the model's default vibe — which is "generic AI artistic." --sref is what makes output yours.
Look at your prompt. Is there a --sref? If no: this is almost certainly your issue.
If --sref is present: what is the URL or code referencing? A "vaguely cool" Pinterest image, or an actual brand-aligned reference?
Check the --sw weight. Default 100 is weak influence. Try 200-300 for stronger brand inheritance.
Test fix: pick a strong brand-aligned reference image. Add at the start of your prompt with --sw 200. Regenerate. Compare.
If output now looks on-brand, you found the issue. Add --sref discipline to your workflow.
Step 3
Generic prompts produce generic output. "A woman working" gets you stock-photo woman. "A 30-year-old marketing manager in a navy blazer working at a sunlit desk" gets you a real person.
Read your prompt aloud. Could it describe 1,000 different images? Then it is too generic.
Specific subjects: age, clothing, demeanor, action. "A focused 32-year-old founder typing on a MacBook" > "A person working."
Specific settings: time of day, light type, room details. "Sunlit corner office with houseplants, late afternoon" > "An office."
Specific styles: name a photographer or film stock. "Shot on Kodak Portra 400, editorial style, Annie Leibovitz mood" > "Professional photo."
The trade-off: more specific = less variety in batch. That is the right trade for production work where you want intentional output.
Step 4
Old V5 tutorials taught "8k ultradetailed cinematic masterpiece" word stacks. In V7 these actively reduce quality.
Look at your prompt for: "8k," "4k," "ultradetailed," "hyperrealistic," "cinematic," "masterpiece," "award-winning," "trending on artstation," "best quality."
These were V5 hacks. V7 interprets them as confusion signals.
Remove all of them. Re-generate.
In their place, use concrete style language: "shot on [film stock]," "[lighting type]," "[specific photographer or style] mood."
Cargo-cult removal often improves output meaningfully on its own.
Step 5
If --v is unset and your defaults are old, you might be on V5/V6. Modern brand-quality output requires V7.
Check the prompt. Is --v 7 explicit? If not, your default version applies.
Open Midjourney settings → Default model version. Should be V7 (current).
If on V6 or older: switch to V7. V6 was great in 2024 but V7 is meaningfully better at photorealism and prompt adherence.
Niji 6 is a different track for anime/manga style. Default V7 for everything else.
Step 6
If you are copying prompts from the public Midjourney feed or AI image Twitter, you are reproducing the same generic styles thousands of others use.
Are you using prompts you found in someone's "Top 10 Midjourney Prompts" article?
These prompts produce instantly recognizable "Midjourney style" — generic by definition because thousands use them.
Original prompt writing + your own --sref library is what produces output that does not look like everyone else's.
Treat public prompt libraries as inspiration for structure, never copy-paste verbatim.
Step 7
After identifying the likely cause, change ONE thing and re-generate. Compare before/after directly.
Pick ONE diagnosis from steps 1-6.
Make exactly that change to your prompt.
Re-generate the same subject and setting.
Compare the new batch to your old batch side by side.
If meaningful improvement: that was the issue. Lock it into your workflow.
If no improvement: revert, try the next diagnosis. Do not stack fixes — you will lose track of what worked.
Common mistakes
Changing 5 things at once and not knowing which fix worked
What goes wrong: You modify stylize, add --sref, rewrite the prompt, change version, all in one go. Output improves but you have no idea why. Next time you cannot reproduce the fix.
How to avoid: Change ONE variable, re-generate, compare. Repeat. This is how you learn what your prompts respond to.
Blaming Midjourney for what is a brief problem
What goes wrong: Your prompt is generic ("a tech founder") because your brief was generic. Output reflects the brief. You blame the tool and switch to DALL-E, where the same generic brief produces equally generic output.
How to avoid: Sharpen the brief first. Specific subject, specific setting, specific style. Then prompt.
Using --p (personalization) trained on bad inputs
What goes wrong: You enabled --p with 30 random clicks early on. Now every generation drifts toward your personalization, which is essentially noise. Output feels "weird" not "branded."
How to avoid: Settings → reset Personalization. Or disable --p. Curate 200+ deliberate likes before re-enabling.
Keeping --chaos high during production
What goes wrong: High chaos (50+) gives wildly different outputs per batch. You cannot iterate; each batch is a fresh exploration. Production stalls.
How to avoid: Drop --chaos to 0-10 once you have found a direction. Chaos is for exploration, not production.
Ignoring aspect ratio mismatch
What goes wrong: Generating at default 1:1 then cropping to 9:16 for a story. The output was composed for square; cropped it loses the subject. Looks generic and badly framed.
How to avoid: Always generate at the target aspect ratio. Per-platform --ar discipline.
Skipping Vary in favor of re-prompting
What goes wrong: You get a 70% usable image, change the prompt, lose what worked. New batch is generic. You blame the tool.
How to avoid: Use Vary (Subtle) or Vary (Region) to iterate. Preserve what works; only modify what does not.
Recap
Done — what's next
Midjourney prompt engineering basics (V7, 2026)
Read the next tutorial
Hand it off
Generic output is usually a workflow problem, not a prompt problem. A specialist diagnoses the root cause in 30-60 minutes and rebuilds your prompt system in one engagement. EverestX matches in 48 hours, $14-16/hr — most diagnostic + fix engagements run $80-200.
See specialist rates
Three usual reasons: (1) you are using public prompt templates everyone else uses. (2) No --sref, so output defaults to Midjourney's generic vibe. (3) Cargo-cult quality words ("8k ultradetailed") that produce the recognizable "Midjourney slop" look. Fix all three.
20-35 descriptive words for production work. Specific subject (age, clothing, action), specific setting (time, light, location details), specific style (named photographer or film stock). More than 40 words averages into mush; fewer than 15 produces generic results.
Only after you have curated 200+ deliberate likes. Turning on --p too early (with random exploration clicks) actively hurts output by training the model on your "noise" rather than your "taste."
The issue is probably prompt specificity or your brief. Generic in = generic out. Re-write the prompt with specific subject, setting, and style. Or get a specialist to rebuild your prompt system in one session.
For photorealism and prompt adherence, yes. For specific illustrated/painted styles, V6 sometimes has more character. For 95% of marketing/brand use cases, default V7. Only use --v 6 explicitly when you have an A/B test showing it is better for a specific look.
Midjourney
Most "ultimate Midjourney prompt guides" are V5-era cargo cult. This walks through the V7-native prompt structure professionals actually use — and the prompt patterns that consistently produce on-brand commercial imagery.
Midjourney
--sref is the difference between "I got a great image once" and "I can ship 100 on-brand images per campaign." This is the workflow that production AI image teams actually use.
Midjourney
Brand consistency in Midjourney is not luck. It is a system: SREF library, prompt templates, color/lighting rules, and a QA pass. This is the production system pros use.
Midjourney
DIY Midjourney is a great idea — until it is not. This is the honest framework: when the cost of self-managing your AI imagery workflow exceeds the cost of hiring help, and what to look for in a specialist.