Midjourney Prompt Guide for Stunning AI Art in 2026

Midjourney V7 is the strongest aesthetic image model on the market in 2026. But "best model" doesn't matter if you can't prompt it. This guide gives you the 9-element structure, every parameter that actually moves the needle, 30+ tested prompts across the 12 aesthetics going viral right now, and the mistakes that quietly produce mediocre output.

Why Midjourney V7 specifically

V7 (released late 2025) shipped three changes that rewrote the prompt playbook: predictable style stacking, vastly improved hands and typography, and the new --style raw mode that gets you photographic fidelity without losing Midjourney's signature aesthetic. If you learnt to prompt on V5 or V6, expect a 30% quality jump from re-learning the basics.

The 9-element prompt structure

Every prompt that consistently produces shippable output follows this skeleton, in order:

1. Shot type — wide / medium / close-up / extreme close-up / over-the-shoulder
2. Subject — one subject, 4-6 words. More subjects = mangled faces.
3. Action verb — specific motion: "striding", "exhaling", "tilting head"
4. Location — concrete, not generic: "Marrakesh souk at dusk" beats "a market"
5. Lighting — golden hour, soft top-light, neon side-light, candle-lit
6. Lens / film — 35mm Kodak Portra, anamorphic 50mm, IMAX 70mm
7. Camera motion or mood — locked off, slow push-in, contemplative
8. Style modifier — "in the style of [movement, not artist]"
9. Parameters — --ar, --s, --style, --chaos

Worked example

Medium close-up of a Japanese chef plating sushi at dusk, hands moving with surgical precision, soft top-light from a paper lantern, 35mm Kodak Portra grain, contemplative mood, quiet-luxury editorial --style raw --s 200 --ar 3:4

Three usable frames out of three. The boring version — "a chef making sushi" — gives you generic stock-photo blur.

The parameters that actually matter

--ar (aspect ratio)

• 3:4 — vertical editorial, magazine spreads, Pinterest pins
• 4:5 — Instagram feed default
• 9:16 — Stories, Reels, TikTok
• 3:2 — landscape editorial, classic 35mm
• 2:3 — portrait classic
• 16:9 — YouTube thumbnails, hero banners
• 2.39:1 — cinematic wide screen

--s (stylize)

Controls how much Midjourney's aesthetic layer overrides your prompt.

• --s 50 — close to prompt, less stylised. Good for photography.
• --s 100 — default. Balanced.
• --s 250 — strongly stylised. Good for illustration and fantasy.
• --s 500-750 — heavy stylisation. Often too much, but useful occasionally.

--style raw vs default

• --style raw — strips Midjourney's aesthetic layer. Use for photography, product shots, brand work where you want literal interpretation of your prompt.
• Default (no flag) — adds Midjourney's signature look. Use for illustration, conceptual art, anything where aesthetic matters more than fidelity.

--chaos

Controls variation between the 4 outputs. --chaos 0 (default) gives you 4 similar interpretations; --chaos 100 gives you 4 wildly different ones. Use chaos when you're exploring; turn it off when you've found your direction.

--no

Negative prompts. --no blur, --no people, --no text. Surprisingly effective. Use freely.

--cref and --sref

--cref [URL] locks a character's identity across multiple prompts. --sref [URL] locks a style. Both are non-negotiable for serial visual content (a book of illustrations, a brand campaign, an editorial series).

The 12 aesthetics going viral in 2026

1. Cinematic neon noir

Cinematic neon-noir portrait, lone figure standing at a rain-slicked intersection, neon side-light, wet asphalt reflections, anamorphic 50mm, shallow depth of field, contemplative mood --style raw --s 300 --ar 2:3

2. Quiet luxury editorial

Quiet luxury editorial of a linen-clad woman by an open Italian window, muted oat-and-stone palette, natural window light, 35mm Kodak Portra grain, restraint --style raw --s 100 --ar 3:4

3. ISO minimalism

Single product centered, vast negative space, hard directional rim light, deep shadow, brutalist composition, museum-quality lighting --style raw --s 50 --ar 1:1

4. Tokyo glitch

Tokyo glitch portrait, VHS scanlines, CRT phosphor glow, hand-drawn katakana overlay, magenta-cyan duotone, deliberate analog imperfection --s 400 --ar 9:16

5. Cottage baroque

Cottage baroque still life, oil-painting texture, candle-lit, velvet drapery, Old Masters chiaroscuro, organic forms, wellness-aesthetic palette --s 350 --ar 4:5

6. Brutalist product

Brutalist product photography of a black ceramic vessel, monolithic concrete shadow, single hard light source, geometric simplicity, premium DTC editorial --style raw --s 50 --ar 4:5

7. Soft 3D claymation

Soft 3D claymation character, rounded pastel forms, gentle imperfect surfaces, Pixar-meets-Behance, golden-hour studio light --s 350 --ar 1:1

8. Vaporwave revival

Vaporwave revival aesthetic, sun-faded gradient palette, Roman bust statue, palm shadows, magenta-cyan duotone, 1989 mall poster energy --s 400 --ar 3:4

9. Craft paper texture

Layered craft-paper illustration, hand-cut shapes with raw edges, drop shadows, editorial children's-book aesthetic, warm earthy palette --s 300 --ar 4:5

10. Cyber-pastoral

Cyber-pastoral scene, satellite dish next to a thatched-roof barn at dusk, eerie quiet, telephoto compression, dystopian-but-beautiful --style raw --s 150 --ar 16:9

11. Anti-aesthetic

Anti-aesthetic snapshot, blown-out flash, point-and-shoot camera, deliberately ugly framing, Gen-Z editorial energy --style raw --s 25 --ar 4:5

12. Museum archival

Museum archival photograph, daguerreotype texture, archival fade, sepia tones, century-old emotional gravity --style raw --s 100 --ar 3:4

A working creator's workflow

1. Pick three aesthetics from the 12 above that fit your brand. Don't try all twelve.
2. Build one prompt template per aesthetic with 4-6 swappable variables.
3. Generate 4 variations per template per day for two weeks.
4. Curate the strongest 10 outputs into a prompt pack.
5. List the pack on OBSYNK with cinematic preview images.
6. Write a long-form article in the OBSYNK Journal teaching the aesthetic. Drives compounding SEO + sales.

Mistakes that quietly produce mediocre output

• Too many subjects. "A group of people" mangles faces. Use "lone figure", "three figures", or specific named subjects.
• Empty adjectives. "Beautiful, stunning, gorgeous" — meaningless to V7. Describe the lighting and the composition.
• Director name-dropping. Sora 2 blocks it; Midjourney mostly ignores it. Describe the look directly.
• Skipping --ar. Default 1:1 is usually wrong for your medium. Always specify.
• Using --s 750 by default. Heavy stylisation is rarely what you actually want. Start at 100-250.
• Not using --sref. If you want consistent visual identity across a series, --sref is non-negotiable.

Advanced moves

Style stacking

V7 finally handles multiple style modifiers cleanly. "Quiet luxury editorial, brutalist composition, Kodak Portra grain" blends three signals reliably.

Image weighting (--iw)

Adjust how much of an input image's identity Midjourney preserves. --iw 2 is high fidelity; --iw 0.5 uses the image only as inspiration.

Vary (subtle / strong)

Don't re-prompt — vary. The Vary buttons give you tighter iteration than starting from scratch.

Browse 12 aesthetic-specific Midjourney prompt packs at OBSYNK Explore — every pack is parameter-tested, with preview images and use-case notes included.

Five less obvious mistakes that quietly tank Midjourney output

Beyond the headline mistakes, five subtler errors creators make once they think they've mastered the basics:

1. Adjective stacking without a structural role

"Beautiful, ethereal, dreamy, magical, stunning, breathtaking, mystical, otherworldly portrait" — the model averages these out and gives you mush. Each adjective should play a structural role (lighting, mood, lens) or it shouldn't be there.

2. Over-relying on "in the style of [movement]"

"In the style of art deco" works. "In the style of art deco meets cyberpunk meets minimalism" produces visual confusion. Stack at most two style modifiers, and make sure they have a coherent reason to coexist.

3. Ignoring color language

"Warm" is fine. "Sodium-vapour orange with cyan deep-shadow accents" is far more controllable. Learn 20-30 specific color descriptors. They give you the most direct control of mood with the fewest words.

4. Generic camera language

"Cinematic" doesn't tell Midjourney what cinema you mean. "Roger Deakins lighting" is filtered. But "soft top-light, 1.85:1 framing, gentle dolly-push, Kodak Vision3 grain" gives you specific control that compounds.

5. Not using image references

The single biggest jump in V7 output quality comes from feeding in 1-2 reference images via image prompts or --sref. Even a vague mood-board image dramatically tightens the model's interpretation. Most creators still don't use this regularly.

Five genre-specific Midjourney playbooks

Editorial fashion

Editorial fashion portrait, lone female figure in oversized linen, oat-and-stone palette, natural window light from camera-left, 35mm Kodak Portra grain, half-body composition with negative space on the right, magazine-grade restraint --style raw --s 100 --ar 3:4 --sref [optional reference URL]

DTC product photography

Premium DTC product photography of a [PRODUCT], centred composition, monolithic concrete shadow, single hard rim-light from upper-right, deep shadow, brutalist negative space, museum-grade lighting precision --style raw --s 50 --ar 4:5

Real-estate listing image

Real-estate listing photograph of a modern Mumbai apartment living room at golden hour, large floor-to-ceiling windows showing the city beyond, mid-century furniture, warm sodium-vapour reflections, 16mm wide-angle, two-point perspective --style raw --s 50 --ar 16:9

Children's book illustration

Hand-cut craft-paper illustration of a small fox crossing a moonlit forest, layered paper textures with raw edges, drop shadows, warm earthy palette, children's-book editorial aesthetic --s 350 --ar 4:5

Cinematic concept art

Cinematic concept art, lone protagonist silhouetted against an oncoming dust storm in the high desert, blown-out warm-amber lighting, anamorphic 2.39:1, painterly atmospheric haze, Roger Deakins-grade composition --s 250 --ar 2.39:1

A working creator's Midjourney week

The cadence top OBSYNK creators run, week after week:

• Monday: generate 40-60 outputs across 3 prompt templates. Pick the 5 best.
• Tuesday: refine the 5 picks (Vary buttons, --iw tuning, --s adjustments). Lock the 3 strongest.
• Wednesday: upload to OBSYNK with cinematic preview images and proper titles.
• Thursday: publish process content — one before/after carousel on Instagram + LinkedIn + Pinterest.
• Friday: write a short OBSYNK Journal post about the week's technique. Builds SEO authority compounding over months.

Five outputs per week is realistic, recognisable, and sustainable. Most creators try to publish 30 per week and burn out by week six.

If you want to go from generating to selling

The single highest-leverage move once your prompts consistently produce shippable output: package them. Three to five prompts in a coherent style, listed as a pack on OBSYNK at ₹999-₹2,999, generates ₹50k-₹3L/month for top creators. The platform's discovery surface (Pinterest-style trending feed, category pages indexed for SEO and AEO) does the audience-building so you can focus on craft.