Midjourney V7: The Complete Prompting Guide

Welcome: Why V7 Changes Everything

If you've been using Midjourney for a while, you've built up muscle memory. You know the shortcuts: pile on the adjectives, include "8K ultra-detailed masterpiece," throw in "cinematic lighting, award-winning photography" for good measure. That approach worked. Sort of. In V5 and even V6, padding your prompts with quality buzzwords nudged the model toward better results.

V7 breaks all of that. Not gently — it actively penalizes the old way.

Midjourney V7, released April 3, 2025 (and made the default model on June 17, 2025), represents the most significant architecture shift in Midjourney's history. The model was retrained from the ground up with a fundamentally different philosophy: understand language the way a human does, not the way a keyword search engine does.

Think about how you'd describe a photograph to a talented cinematographer friend. You wouldn't say "stunning, gorgeous, 8K, ultra-detailed, masterpiece." You'd say "It's a jazz musician in his seventies, deep wrinkles, silver stubble, playing saxophone under a single warm spotlight in a basement club — that kind of grainy, intimate Tri-X film look." That's exactly what V7 wants from you.

What's New and Different in V7

• Natural language understanding: V7 parses grammar, sentence structure, and context. Complete sentences outperform comma-separated tags.
• Front-loading matters: The first words of your prompt carry the most weight. Your subject goes first — always.
• New --exp parameter: A V7-exclusive enhancement that adds HDR-like dynamics and detail without blowing up your prompt accuracy.
• --oref replaces --cref: The new Omni Reference system handles subject/identity transfer far more powerfully than the old Character Reference.
• Personalization on by default: V7 uses your taste profile automatically unless you turn it off.
• Dramatically better text rendering: Quotation marks in your prompt now reliably produce legible text in images.
• Draft mode: 10× faster generation at half the GPU cost, purpose-built for exploration.
• Video generation: V1 video model launched June 19, 2025, with motion controls, looping, and HD upscaling.
• Niji 7: The anime/manga model got a major upgrade — more literal, cleaner linework, and now supports personalization (as of February 26, 2026).

The shift from V6 to V7 is the shift from talking to a keyword matcher to talking to a collaborator who actually listens. This guide will show you exactly how to do that.

Quick Start: 3 Copy-Paste Prompts

Not ready to read the whole guide? Start here. These three prompts showcase what V7 does best. Copy them directly into Discord's /imagine command or the Midjourney web app.

1. Cinematic Photorealism

/imagine A quiet fishing village at 5 AM, the harbor mist still heavy, wooden boats reflected in glassy water, a lone fisherman in orange waders preparing nets under a single dock lamp. Long exposure, 35mm lens, muted blues and warm amber tones. --ar 16:9 --style raw --s 50 --q 2

Why it works: Specific scene, clear subject, lighting built into the narrative, --style raw for photographic literalism.

2. Vibrant Concept Art

/imagine An ancient library carved inside a living crystal geode, thousands of glowing manuscripts suspended mid-air between amethyst columns, a robed archivist reaching for a floating tome. The light refracts into prismatic patterns across the stone floor. Detailed fantasy illustration, painterly textures, epic scale. --ar 21:9 --s 400 --exp 15 --c 20

Why it works: Scene and subject together, atmospheric detail, --s 400 for strong artistic interpretation, --exp 15 for enhanced visual richness.

3. Editorial Portrait

/imagine Portrait of an elderly Peruvian weaver, deep laugh lines, silver-streaked black hair pulled back loosely, wearing a handwoven alpaca shawl in deep crimson and saffron. She's looking slightly off-camera with quiet dignity. Natural window light from the left, medium format documentary photography. --ar 3:4 --style raw --s 30 --q 2

Why it works: Rich subject description leads, specific lighting source, the phrase "quiet dignity" communicates mood without generic buzzwords.

Table of Contents

1. Model Identity
2. Prompt Syntax
3. Complete V7 Parameter Reference
4. V7 Prompting Best Practices
5. Word Weighting (Multi-Prompt)
6. Modes: Standard vs. Raw
7. Draft Mode Workflow
8. Reference Systems
9. Personalization and Moodboards
10. Text Rendering
11. Niji 7 (Anime/Manga Model)
12. Video Generation
13. Exp Parameter Deep Dive
14. V6 to V7 Migration Guide
15. Common Parameter Combinations
16. Parameter Recipes
17. Example Prompts (Production-Ready)
18. Prompt Building Workshop
19. Common Mistakes → Better Version
20. Common Pitfalls
21. Cheat Sheet

1. Model Identity

2. Prompt Syntax

/imagine [image URLs (optional)] [text prompt] [--parameters]

Rules:

• Parameters go at the END of the prompt, after all descriptive text.
• Add a space between prompt text and double dashes (--).
• No punctuation inside parameters (no commas, periods).
• No prompt text after parameters.
• Image URLs go immediately after /imagine, before the text prompt.

Example of correct structure:

/imagine A red fox sitting in autumn leaves, dappled forest sunlight, nature photography --ar 3:2 --style raw --s 50

Incorrect (parameters broken up):

/imagine --ar 3:2 A red fox --style raw sitting in autumn leaves

3. Complete V7 Parameter Reference

Core Parameters

Plain-Language Explanations

--s (Stylize) — "How much creative license does Midjourney get?"

This is the most important dial for controlling the overall look. At low values (0–50), Midjourney stays very close to your literal description — what you say is what you get, with minimal added flair. At the default of 100, it adds a moderate aesthetic polish. Crank it above 300 and Midjourney starts making bold artistic choices — unusual color grades, dramatic compositions, stylized textures — that go well beyond your words.

Try this:

• --s 0 — Maximum literalism. Great for technical diagrams or when you need exact control.
• --s 100 — The default. A good balance for most creative work.
• --s 400 — Strong artistic interpretation. Excellent for concept art, fantasy scenes, anything you want to look "designed."
• --s 750 — Midjourney goes full-creative. Often surprising and beautiful, occasionally off-brief.

--c (Chaos) — "How different should each of the 4 images be from each other?"

By default (0), all four images in your grid are variations on the same interpretation. Chaos introduces divergence. At --c 50, you get four meaningfully different takes on your prompt — great for exploring what's possible. At --c 100, the four images might feel almost unrelated.

Try this:

• --c 0 — Consistent variations, useful when you know what you want and just need refinement options.
• --c 25 — Slight variation, keeps the feel consistent but offers some visual diversity.
• --c 75 — High variation. Good for early brainstorming when you don't know which direction to go.
• --c 100 — Maximum surprise. Treat this as a creative lottery.

--w (Weird) — "How strange and unconventional do you want this to get?"

Weird pushes Midjourney toward non-standard aesthetics — unusual proportions, unexpected material choices, surreal combinations. It's not the same as chaos (which varies between images); it shifts the entire aesthetic baseline toward the unusual.

Try this:

• --w 100 — A subtle tilt into the uncanny. Colors might feel slightly off, proportions slightly dreamlike.
• --w 500 — Noticeably strange. Objects might blend, perspectives might warp.
• --w 1500 — Very unusual. Expect unexpected material treatments and surreal logic.
• --w 3000 — Full surrealism. Results are unpredictable but often fascinating.

--exp (Experimental/Enhance) — "Add HDR-like depth and visual richness" (V7-exclusive)

This parameter enhances contrast, detail, and dynamic range in a way that feels like a tone-mapped HDR photograph — more texture in shadows, more depth in highlights, more pop overall. It's subtle at low values and intense at high ones. See Section 13 for the full breakdown.

--q (Quality) — "Spend more GPU time for finer detail"

At --q 1 (default), you get a good result. At --q 2, the model spends twice the compute on fine details — textures get crisper, edges cleaner, subtle gradients smoother. At --q 4, four times the compute. The gains from --q 4 over --q 2 are often marginal for most subjects but noticeable in highly detailed scenes.

Try this:

• --q 1 — Default. Use for exploration, drafts, and most everyday generation.
• --q 2 — Final renders, portraits, product shots, anywhere fine detail matters.
• --q 4 — Architecture, intricate patterns, detailed landscapes where every pixel counts.

--seed — "Lock in a starting point for reproducibility"

Every image generation starts from a random noise seed. By specifying --seed 12345, you always start from the same noise, which means very similar (though not identical) images if your prompt stays the same. Invaluable when you've found a composition you love and want to iterate on it without losing the core feel.

Style and Reference Parameters

Plain-Language Explanations

--oref (Omni Reference) — "Keep this person/character/subject consistent across images"

This is V7's answer to the challenge of character consistency. You provide a reference image URL and Midjourney transfers the visual identity — face, body type, clothing, distinctive features — to your new image. This replaced --cref from V6 and is significantly more capable.

Try this:

• --oref [URL] --ow 60 — Loose identity transfer, great when you want the character to feel related but adapted to the new scene's style.
• --oref [URL] --ow 150 — Strong identity lock. The character's face and key features will stay very close to the reference.
• --oref [URL] --ow 300 — Maximum fidelity. Use when exact likeness matters more than aesthetic flexibility.

--sref (Style Reference) — "Make this image look like that image"

Provide a reference image and Midjourney borrows its visual language: color palette, lighting quality, brushwork or photographic style, compositional mood. You can stack multiple reference URLs to blend styles.

Try this:

• --sref [URL] --sw 50 — A whisper of the reference style underneath your prompt's own aesthetic.
• --sref [URL] --sw 200 — The style clearly comes through without completely overriding your prompt.
• --sref [url1] [url2] --sw 150 — A blend of two aesthetic references.

--style raw — "Literal mode. No AI embellishment."

By default, Midjourney adds its own aesthetic flair — a certain smoothness, a tendency toward beautiful compositions, a slight "AI painting" quality. Raw mode turns that off. The model becomes more literal and photographic. Essential for product photography, technical reference images, and photorealism where you need the output to look like a real photograph rather than a beautiful AI painting of one.

Speed/Mode Parameters

Plain-language: Turbo is great when time matters and you're flush with fast hours. Fast is the everyday default. Relax is for bulk generation when you're not in a hurry. Draft is the hidden gem — use it to rapidly explore dozens of compositional directions at minimal cost before committing to a full-quality render.

Video Parameters

Model Selection

4. V7 Prompting Best Practices

4.1 The Fundamental Shift

V7 fundamentally changed how prompts work. The old keyword-soup approach actively degrades your results. V7 understands natural language — write prompts like describing a photograph to a skilled cinematographer.

BAD (V5/V6 style — DO NOT USE):

beautiful, stunning, 8k, detailed, masterpiece, photorealistic, cinematic lighting, award-winning

GOOD (V7 natural language):

A weathered lighthouse keeper standing at the top of a spiral staircase, golden afternoon light streaming through salt-stained windows, documentary photography style

The difference isn't just stylistic preference — the keyword-soup approach produces measurably worse results in V7 because the model is trying to reconcile a list of disconnected aesthetic demands rather than building a coherent visual scene.

4.2 Prompt Structure Template

[SUBJECT] [SUBJECT DETAILS], [CONTEXT], [STYLE/MOOD], [TECHNICAL] --parameters

4.3 Critical Rules

1. Front-load the subject. Early words carry the most weight.

• BAD: "shot on film, photorealistic, woman in red dress"
• GOOD: "Woman in red dress, photorealistic, shot on film"

1. Use complete sentences with proper grammar. V7 parses grammar, spelling, and punctuation for meaning.
2. Be specific. "1970s animated cartoon style" outperforms "cartoon".
3. Use redundancy for emphasis. If Midjourney ignores a detail, describe it multiple ways:

• "a mechanical parrot with visible gears, brass cogs, and clockwork mechanisms"

1. Optimal token count: 50–150. Under 30 = open interpretation. 150+ = diminishing returns.
2. Avoid contradictions. "dark, bright, moody, cheerful" confuses the model.
3. Quality signals that matter in V7:

• Lighting descriptions (most impactful)
• Material/texture specifications
• Atmosphere/mood descriptors
• Technical camera terms

1. Quality signals that DO NOT help in V7:

• "8K", "ultra-detailed", "masterpiece", "award-winning" — ignored or degrading

4.4 Writing Great Lighting Descriptions

Lighting is the single highest-impact element in your prompt. Here are phrases that actually work:

• "single warm spotlight from above" — dramatic portrait lighting
• "golden hour, sun at the horizon, long shadows" — warm, directional outdoor light
• "overcast diffused light, no harsh shadows" — even, soft, documentary feel
• "neon sign reflections on wet pavement" — urban night scene
• "candle-lit interior, warm pools of amber light" — intimate, historical
• "rim-lit from behind, silhouette at the edge" — cinematic drama
• "north-facing studio window, soft indirect daylight" — classic portrait/still life
• "bioluminescent glow from below" — sci-fi or fantasy mood

4.5 Writing Great Texture and Material Descriptions

Materials give V7 enormous amounts of visual information:

• "brushed aluminum with fine horizontal striations"
• "hand-thrown ceramic with visible finger marks in the glaze"
• "weathered oak with deep grain and silver patina"
• "matte carbon fiber, hexagonal weave pattern"
• "poured concrete with aggregate visible at the surface"
• "liquid mercury pooling on black glass"

5. Word Weighting (Multi-Prompt)

V7 supports explicit weight control using :: syntax:

When to use :: weighting vs --no:

Use :: weighting when you want to push the balance between two present elements — for example, you want both a knight and a dragon but want the dragon to dominate compositionally. Use --no when you simply want to exclude something from the image entirely.

Example:

/imagine moonlit castle::1 ancient dragon::3, the dragon perched on the highest tower, dramatic storm clouds behind it --ar 16:9 --s 350

The dragon gets three times the compositional weight of the castle setting.

6. Modes: Standard vs. Raw

Pro tip: For maximum photorealism: --style raw --s 50

Choosing between modes:

• Use Standard when you want Midjourney to make the image look "beautiful" or "designed" — it'll add compositional polish and aesthetic cohesion that often elevates creative work.
• Use Raw when you need the output to look like a real photograph, a specific technical illustration, or when you're providing very detailed artistic direction and don't want Midjourney second-guessing you.
• Try the same prompt in both modes — the contrast is often illuminating.

7. Draft Mode Workflow

Recommended workflow:

1. Generate in Draft Mode (--draft) to explore compositions, test variations
2. Identify winning direction
3. Re-generate without --draft for full quality
4. Use Enhance button to upscale Draft results

Draft mode is underused. Most experienced users run 5–10 draft generations for every full-quality render. The compositional work happens in draft; the polish happens in full. At 10× speed and half the cost, you can explore far more creative territory before committing.

Draft + Chaos is a powerful exploration combo:

/imagine A futuristic market bazaar on a space station --draft --c 75 --ar 16:9

Run this five times to quickly survey a wide range of compositional approaches, then pick the most promising and re-generate at full quality with the same seed.

8. Reference Systems

8.1 Omni Reference (--oref) — Subject Identity

Replaces --cref from V6. Transfers subject characteristics: face, body, clothing, identity.

Usage:

/imagine [prompt] --oref [image URL] --ow 100

Best practices:

• Use high-resolution, clear reference images
• Front-facing for faces
• Consistent lighting
• Minimal/clean background
• Increase --ow when using high --s or --exp (they compete)

NOT supported in Niji 7.

Advanced tips:

• For fashion/clothing consistency (not face): use a reference image focused on the outfit rather than the face, and keep --ow moderate (60–100).
• For character consistency across art styles, try --oref [URL] --ow 80 --sref [style URL] --sw 200 to maintain the character while adapting the style.
• If the reference doesn't look quite right, try the same prompt with a slightly different --ow value — the relationship between --ow, --s, and --exp is nonlinear.

8.2 Style Reference (--sref) — Aesthetic Transfer

Transfers visual aesthetic: color palette, lighting quality, technique, mood, composition.

Usage:

/imagine [prompt] --sref [image URL] --sw 150

Multiple styles (blending):

/imagine [prompt] --sref [url1] [url2] --sw 200

Legacy V6 style codes: Use --sv 4 or --sv 6 to apply older codes.

Practical style reference strategies:

• Painter as style reference: Find a high-quality image of a painting in your preferred style and reference it. The model will borrow technique, palette, and mood without copying the content.
• Film still as style reference: A frame from your favorite film transfers its cinematographic language — color grading, lighting ratio, lens characteristics — to your entirely different subject.
• Product photography as style reference: Reference a commercial photo with excellent lighting to transfer that professional product photography look.
• Blend complementary styles: --sref [impressionist painting] [neon photography] --sw 180 creates a fascinating photographic-impressionist hybrid.

8.3 Combining References

/imagine [prompt] --oref [subject image] --sref [style image] --ow 100 --sw 150

This combination — consistent subject (--oref) with a specific visual style (--sref) — is one of V7's most powerful features. It enables genuine character consistency across different visual styles, something that was extremely difficult in previous versions.

9. Personalization and Moodboards

Personalization

• Enabled by default in V7.
• Based on image-pair rankings (200+ ratings recommended for stability).
• Toggle: --p (on with default profile), --p [mID] (specific profile).
• To disable: adjust in settings or use --p 0.

Moodboards

• Custom style profiles created by rating images at midjourney.com/personalize.
• Stability tiers: 40 ratings = minimum usable, 200 = fairly stable, 2000 = maximum refinement.
• Share moodboards as codes: --profile 2jrqbw6
• Blend with style references: --p --sref [url] --sw 50

Building a useful moodboard: Start by rating 200+ image pairs at midjourney.com/personalize. Focus on being decisive — don't overthink each choice. Your instinctive reactions are what the profile is capturing. After 200 ratings, your --p profile will consistently push your generations toward your aesthetic sensibility, working almost like a background style guide for everything you create.

Moodboard strategy for professionals: Create separate moodboards for different client aesthetics or project types, then switch between them using --profile [mID]. This lets you quickly shift from, say, a clean corporate photography style to a gritty editorial documentary style by simply changing the profile code.

10. Text Rendering

V7 dramatically improved text rendering. Use quotation marks:

A vintage movie poster with the title "MIDNIGHT EXPRESS" in bold Art Deco lettering at the top

For best results:

• Keep text short (1-3 words ideal)
• Specify font style or reference
• Place text description naturally in the prompt

Font style vocabulary that works well in V7:

• "bold Art Deco lettering" — geometric, 1920s–30s aesthetic
• "handwritten chalk script" — casual, organic
• "engraved serif typeface" — formal, traditional
• "neon tube lettering" — glowing, urban
• "gothic blackletter" — medieval, heavy
• "thin geometric sans-serif" — modern, minimal
• "distressed 1970s display type" — retro, grungy

Practical text tip: For logos or signage where text accuracy is critical, generate with --q 2 and use --no blurry text, distorted letters in your negative prompt. Longer text strings (5+ words) remain unreliable — use V7 for text accents, not body copy.

11. Niji 7 (Anime/Manga Model)

Usage: --niji 7

Niji 7 tips:

• More literal than previous Niji versions — describe styles explicitly rather than relying on vibes
• No style presets — describe the anime style you want directly
• Cleaner linework, improved eye/reflection detail
• Best --sref consistency of any model

Niji 7 style vocabulary:

Previous Niji models let you invoke shortcuts like "shonen style" and trust the model's interpretation. Niji 7 is more literal — it needs explicit direction. Here are specific style descriptors that work:

• "Studio Ghibli-esque watercolor backgrounds, soft character outlines"
• "1990s OVA aesthetic, muted colors, heavy crosshatching in shadows"
• "Shounen manga panel, bold ink lines, speed lines, minimal color"
• "Modern manhwa style, webtoon coloring, clean linework"
• "Noir anime, high contrast shadows, rain-soaked streets"
• "Moe aesthetic, soft gradients, oversized luminous eyes"
• "Katsuhiro Otomo-inspired, detailed dystopian backgrounds, realistic proportions"

12. Video Generation

Video is web-only (not Discord). V1 model launched June 19, 2025.

HD Video: 4× sharper, 3.2× cost. Pro/Mega plans only, Fast mode only. Generate SD first, then HD.

Extend: Add +4 seconds up to 4× (21 seconds max). Auto/Manual/Loop modes.

Video generation strategy:

For cinematic shots: Use --motion low — this produces gentle camera movement and subtle scene animation that feels like professional cinematography rather than frantic AI movement.

For dynamic scenes: --motion high works well for action, weather, flowing water, and anything where movement is part of the story. Be cautious with character faces at high motion — they can drift.

For seamless loops: --loop is excellent for ambient backgrounds, looping animations, and social media content. Start with a static or near-static image prompt — complex scenes loop less cleanly.

For morphing between scenes: --end [URL] lets you specify where the video lands, creating a controlled transformation. The model interpolates between your starting prompt and the end frame image.

Example video prompts:

/imagine A lone candle flame in a dark room, ultra close-up, the flame swaying gently in an unseen draft. --motion low --loop --ar 9:16

/imagine A thunderstorm approaching over open ocean, dark cumulonimbus clouds building, whitecaps forming on choppy water, dramatic light. --motion high --ar 16:9

/imagine Cherry blossom petals drifting past a lantern-lit garden path at night. --motion low --loop --ar 3:4

13. Exp Parameter Deep Dive

--exp is V7-exclusive. It adds enhanced detail, dynamics, and tone-mapped HDR-like effects.

Warning: High --exp (>25-50) competes with --s and --p, reducing their influence. For combinations:

• --exp 10 --s 200 — balanced
• --exp 25 --s 100 — exp-forward
• --exp 5 --style raw — subtle enhancement on raw output

What does --exp actually look like?

At --exp 10, you'll notice: richer shadow detail, more specular highlights on surfaces, a slight increase in perceived contrast and depth. It's the difference between a flat photograph and one that's been thoughtfully tone-mapped — the scene gains a three-dimensional quality.

At --exp 25, the effect becomes dramatic: textures pop, the image has a heightened cinematic quality, colors become more saturated and punchy. It's beautiful for landscapes, architectural photography, and concept art. Be careful with portraits — it can over-dramatize skin texture.

At --exp 50+, the effect starts dominating the entire image. Your prompt's specific instructions can get steamrolled by the model chasing visual drama. Use with care.

--exp use case guide:

• Product photography: --exp 5-10 — adds material richness without over-processing
• Landscape/environment: --exp 15-25 — excellent for dramatic natural scenes
• Portrait: --exp 5-15 — keeps skin natural while enriching background and clothing
• Concept art: --exp 10-25 — adds painterly richness and depth
• Abstract/pattern: --exp 20-50 — can create extraordinary visual complexity

14. V6 to V7 Migration Guide

This section is for users who have been working with V6 and need to adapt their workflows.

The Core Changes

Side-by-Side Prompt Translations

Portrait Photography

V6 style:

beautiful woman, 30s, auburn hair, green eyes, leather jacket, rooftop, golden hour, photorealistic, 8k, detailed, cinematic lighting, bokeh, sharp focus, professional photography

V7 style:

A woman in her early thirties with curly auburn hair and green eyes, wearing a worn leather jacket, standing on a rooftop at golden hour. The city skyline behind her is soft and warm. Shallow depth of field, editorial photography, natural light. --ar 3:4 --style raw --s 50

Fantasy Landscape

V6 style:

fantasy landscape, magical forest, glowing mushrooms, misty, epic, ultra-detailed, concept art, artstation, trending, stunning, dramatic lighting

V7 style:

A primeval forest floor at dusk where oversized bioluminescent mushrooms cast a cool blue-green glow across the moss and fern undergrowth. Mist drifts between ancient tree trunks. Painterly fantasy concept art, rich atmospheric depth. --ar 16:9 --s 350 --exp 15

Product Shot

V6 style:

perfume bottle, luxury, white background, studio lighting, photorealistic, commercial, 8k, perfect, clean

V7 style:

A tall amber glass perfume bottle on a white marble surface with subtle veining. Single overhead key light creating a clean shadow to the lower right. Shot on a medium format camera with a 90mm macro lens. Commercial beauty photography. --ar 4:5 --style raw --s 30 --no reflections, dust, fingerprints

Character Art

V6 style:

warrior woman, armor, sword, forest, detailed, intricate, fantasy, epic, concept art, artstation, digital painting

V7 style:

A warrior woman in battle-worn plate armor, a longsword slung across her back, standing at the treeline of an ancient forest at dawn. Her expression is calm and determined. The rising sun backlights her, casting long shadows through the trees behind her. Epic fantasy concept art, detailed metalwork, painterly. --ar 2:3 --s 400 --exp 10

Architecture

V6 style:

futuristic city, cyberpunk, neon, rain, detailed, ultra-detailed, 8k, cinematic, dramatic lighting, masterpiece

V7 style:

An elevated walkway between two glass skyscrapers in a rain-soaked future city at midnight. Neon signage in Japanese and Chinese reflects in the puddles below. A lone figure in a clear umbrella walks toward us, backlit by a delivery drone's lights. Cinematic wide angle, cyberpunk atmosphere. --ar 21:9 --s 300 --exp 20

Anime Scene

V6 style:

anime girl, school uniform, cherry blossoms, beautiful, detailed eyes, artstation, trending, masterpiece, stunning

V7 style:

A teenage girl in a dark navy school uniform stands alone in a courtyard as cherry blossoms fall around her. She looks up at the sky with an expression of quiet longing. Delicate ink linework, soft watercolor fills, 1990s-era shoujo manga aesthetic. --niji 7 --ar 3:4 --s 200

The One Rule That Changes Everything

In V6, your prompt was a specification list. In V7, your prompt is a scene description.

Ask yourself: if I handed this text to a skilled photographer and said "make this image," would they have enough information? Would they understand the mood, the light, the character, the environment? If yes, your V7 prompt is working. If it reads like a list of desired attributes, rewrite it as a scene.

15. Common Parameter Combinations

Photorealism

--style raw --s 50 --no anime, cartoon, illustration, painting, drawing

Stylized Art

--s 400 --exp 10 --p

High Variation Exploration

--c 50 --draft

Subject Consistency Series

--oref [URL] --ow 150 --sref [style URL] --sw 100

Anime/Manga

--niji 7 --sref [anime reference URL] --sw 200

Seamless Patterns

--tile --style raw --s 50

16. Parameter Recipes

Pre-built parameter combinations for common use cases. These are starting points — adjust to taste.

Commercial Photorealism

Goal: Images that could pass as real photographs, suitable for commercial use.

--style raw --s 30 --q 2 --no cartoon, illustration, painting, artistic, stylized

Notes: Low stylize keeps you close to the literal prompt. Raw mode removes AI's aesthetic interpretation. High quality for fine detail.

Dramatic Landscape Photography

Goal: Epic natural scenes with rich atmospheric depth.

--style raw --s 80 --exp 20 --ar 16:9 --q 2

Notes: Exp 20 adds the tone-mapped quality that makes landscape photography pop. Slightly higher stylize than pure photorealism allows for better compositional choices.

Fantasy Concept Art

Goal: Painterly, detailed fantasy illustrations for games, books, or creative projects.

--s 350 --exp 15 --ar 16:9 --q 2

Notes: High stylize for strong artistic interpretation. Exp 15 adds depth and texture richness. Standard mode (no --style raw) lets Midjourney make strong aesthetic choices.

Editorial Portrait

Goal: Magazine-quality portraits with personality and visual interest.

--style raw --s 60 --q 2 --ar 3:4

Notes: Raw mode keeps it photographic. Slightly higher stylize than pure realism allows for interesting compositional choices.

Product Photography

Goal: Clean, professional product shots for e-commerce or marketing.

--style raw --s 20 --q 2 --no text, logo, watermark, reflection, blur

Notes: Very low stylize for maximum product literalism. Strong negative prompt removes common distractions.

Anime/Manga Illustration

Goal: Polished anime-style character or scene illustration.

--niji 7 --s 200 --sref [reference URL] --sw 150 --q 2

Notes: Niji 7 is the right model. Moderate stylize. Style reference to anchor the anime sub-style.

Abstract Art

Goal: Non-representational, visually striking abstract compositions.

--s 750 --w 500 --exp 30 --c 50 --ar 1:1

Notes: High stylize and weird push Midjourney into non-literal territory. High chaos creates divergent options. Exp adds visual richness.

Seamless Surface Pattern

Goal: Tileable patterns for fabric, wallpaper, or digital backgrounds.

--tile --style raw --s 50 --ar 1:1

Notes: --tile flag is essential. Square ratio for versatility. Moderate stylize. Raw mode gives cleaner tile seams.

Dark/Moody Cinematic

Goal: Film noir, horror, gothic, or generally atmospheric dark scenes.

--s 200 --exp 15 --no bright colors, sunshine, cheerful --ar 16:9

Notes: Moderate stylize with exp for drama. Negative prompt prevents Midjourney's tendency toward bright, clean outputs.

Retro/Vintage

Goal: Images with authentic period aesthetics — film grain, color fading, era-specific visual signatures.

--style raw --s 150 --no modern, digital, sharp, crisp --ar 3:2

Notes: Style raw to prevent AI over-polishing. Moderate stylize for character. Negative prompt blocks contemporary polish. 3:2 ratio mirrors film formats.

Hyperdetailed Illustration

Goal: Dense, intricate illustrations where every corner rewards examination.

--s 250 --q 4 --exp 10 --ar 1:1

Notes: High quality is worth it here. Exp 10 adds textural richness. Moderate stylize allows Midjourney to make strong compositional and detail choices.

Architectural Visualization

Goal: Crisp, realistic architectural renderings.

--style raw --s 50 --q 2 --no people, furniture (if exterior focus), vegetation

Notes: Raw mode for realism. Negative prompt customized based on what the architecture scene needs to be free of.

17. Example Prompts (Production-Ready)

Photorealistic Portrait

Original:

/imagine A weathered jazz musician in his 70s, deep wrinkles and silver stubble, wearing a rumpled linen suit and fedora, playing a tenor saxophone under a single warm spotlight in a smoky basement club. Shallow depth of field, 85mm lens, film grain reminiscent of Tri-X 400 pushed to 1600. --ar 3:4 --style raw --s 30 --q 2

Variation 2 — Environmental Documentary:

/imagine An aging Inuit elder on the Arctic tundra, deep weathered face, narrow eyes bright with intelligence, wearing a fur-lined anorak, looking into the middle distance at a grey sky. Wind-blown snow at ground level, the horizon flat and endless behind him. Documentary photography, natural overcast light, medium format film look. --ar 4:5 --style raw --s 40 --q 2

Variation 3 — Studio Fashion:

/imagine A 22-year-old woman with close-cropped platinum hair and sharp cheekbones, wearing an oversized black turtleneck, photographed against a seamless warm grey backdrop. She's looking directly into the lens with flat affect. High-fashion editorial, minimal styling, Avedon-inspired stark simplicity, large format camera. --ar 2:3 --style raw --s 50 --q 2

Variation 4 — Environmental Portrait:

/imagine A master watchmaker in his 60s, reading glasses pushed up on his forehead, bent over a movement on a watchmaker's bench covered in tiny tools, gears, and magnifying equipment. Workshop light from a desk lamp illuminates his hands and the watch. Intimate documentary photography, narrow depth of field. --ar 3:2 --style raw --s 30 --q 2

Product Photography

Original:

/imagine A sleek matte black wireless earbud case resting on a piece of raw black granite. Single key light from upper right creating a dramatic shadow. One earbud floating slightly above the case. Clean, minimal composition. Shot on Phase One IQ4, 120mm macro lens, f/8. Commercial product photography. --ar 4:5 --s 80 --no text, logo, watermark

Variation 2 — Lifestyle Context:

/imagine A handmade ceramic coffee mug with a matte sage green glaze sitting on a raw oak kitchen counter next to a closed book and a sprig of dried lavender. Morning window light from the left, soft and diffused. Lifestyle product photography for a boutique home goods brand. --ar 4:5 --style raw --s 60 --q 2 --no text, price tags

Variation 3 — Luxury Skincare:

/imagine A frosted glass serum bottle with a gold dropper cap, positioned at a 45-degree angle on a piece of white Calacatta marble. A single petal from a white gardenia rests nearby. Three-quarter overhead lighting with a subtle fill from below. Luxury skincare product photography, clean and aspirational. --ar 4:5 --style raw --s 40 --q 2 --no busy background, clutter, hands

Variation 4 — Food Product:

/imagine A small jar of artisan honey with a wooden dipper resting across the top, a slow golden drip of honey mid-fall. Set against a warm terracotta background with a few dried wildflowers scattered casually around the base. Natural studio lighting, slightly warm. Artisan food product photography. --ar 1:1 --style raw --s 70 --q 2

Concept Art

Original:

/imagine A massive abandoned space station orbiting a gas giant, overgrown with bioluminescent alien vegetation that glows turquoise and amber. Interior corridors visible through shattered hull plates. Hard science fiction, detailed engineering, volumetric god rays. --sref [concept art URL] --sw 200 --ar 21:9 --s 300 --exp 15

Variation 2 — Subterranean City:

/imagine A vast underground city carved into the walls of a natural cavern the size of a mountain range. Thousands of glowing amber windows stacked in terraced tiers along the cavern walls. A central column of bioluminescent organisms forms a natural "sun." Tiny gondolas traverse the dark air between districts. Detailed architectural concept art, epic scale, warm light. --ar 21:9 --s 400 --exp 15 --q 2

Variation 3 — Alien Architecture:

/imagine The ruins of an alien temple on a world with a purple atmosphere and two pale moons visible on the horizon. The architecture is all flowing organic curves, as if grown rather than built, now partially collapsed and overgrown. A human exploration team in environment suits is tiny in the foreground for scale. Hard science fiction concept art, documentary realism meets wonder. --ar 16:9 --s 300 --exp 10 --q 2

Variation 4 — Near-Future Urban:

/imagine A vertical farm district in a mid-21st-century Asian megacity — towers wrapped in green growing panels, hydroponic bridges connecting buildings, drone delivery traffic visible against a hazy afternoon sky. Street level shows a wet market with produce from the towers above. Cinematic urban concept art, hyperdetailed, warm late-day light. --ar 21:9 --s 350 --exp 12 --q 2

Character Consistency Series

Original:

/imagine Portrait of Maya, a 28-year-old woman with curly auburn hair, freckles, green eyes, wearing a vintage leather jacket. She's standing on a rooftop at golden hour, city skyline behind her, wind in her hair. Warm tones, editorial photography. --oref [reference URL] --ow 120 --ar 3:4 --s 100

Variation 2 — Same character, different setting:

/imagine Maya — early 30s, curly auburn hair, freckles, intense green eyes — in a quiet library, sitting cross-legged on a window seat with rain on the glass behind her, reading a thick novel. She glances up at the camera, one corner of her mouth slightly lifted. Soft grey afternoon light. --oref [reference URL] --ow 120 --ar 3:4 --s 80

Variation 3 — Same character, action scene:

/imagine Maya, late 20s with curly auburn hair and freckles, in a full sprint through a rain-soaked alley at night, a messenger bag over her shoulder, motion blur on the background, sharp on her face. Blue-white sodium arc light overhead. Cinematic action photography. --oref [reference URL] --ow 100 --ar 2:3 --s 100 --exp 10

Variation 4 — Same character, fantasy reinterpretation:

/imagine Maya as a medieval herbalist, same curly auburn hair and freckles, in a candlelit apothecary with shelves of glass jars behind her. She's examining a glowing botanical specimen. Fantasy illustration, warm candlelight, detailed. --oref [reference URL] --ow 80 --sref [fantasy art URL] --sw 150 --ar 3:4 --s 250

Anime/Niji 7

Original:

/imagine A solitary samurai standing at the edge of a cliff overlooking a sea of cherry blossoms below, katana drawn, wind whipping their long black hair and tattered white haori. Dramatic backlighting from a blood-red sunset. Detailed ink linework, cel shading, Studio Ghibli meets Vagabond manga aesthetic. --niji 7 --ar 16:9 --s 200

Variation 2 — Slice of Life:

/imagine Two high school students sharing an umbrella at a train platform in the rain, one slightly taller, neither looking at the other, both staring at the arriving train. Warm train windows glow against the grey afternoon. Soft pencil-style line art, muted palette, contemporary shoujo manga aesthetic. --niji 7 --ar 3:4 --s 150

Variation 3 — Mecha/Action:

/imagine A pilot in a scarred cockpit, face lit by warning lights, pulling back on the controls as a giant mecha hand reaches toward the camera. Dramatic forced perspective, motion blur on the hand. Classic 1990s OVA aesthetic, heavy cross-hatching in shadows, limited color palette of red, grey, and black. --niji 7 --ar 16:9 --s 250

Variation 4 — Atmospheric Fantasy:

/imagine A young onmyoji in Heian-period court robes surrounded by paper shikigami she has animated, standing in a moonlit garden. The shikigami take the forms of birds and foxes and glow faintly white. Delicate brushwork, gold and indigo night palette, classical Japanese woodblock meets modern anime illustration. --niji 7 --ar 9:16 --s 200 --exp 5

Variation 5 — Cyberpunk Anime:

/imagine A hacker in a neon-lit capsule hotel room, holographic code cascading around her, cables connecting her headset to an unseen network. Sweating, intense concentration. Gritty cyberpunk anime aesthetic, high contrast, cool blue-magenta palette, Ghost in the Shell visual language. --niji 7 --ar 2:3 --s 300

Seamless Patterns/Tiles

Original (inferred from parameters section):

/imagine A seamless pattern of stylized botanical elements — tropical leaves, hibiscus flowers, and exotic birds in a flat graphic style, warm tropical palette of coral, jade, and ivory --tile --style raw --s 50 --ar 1:1

Variation 2 — Geometric:

/imagine A seamless geometric pattern of interlocking hexagons at three different scales, rendered in a deep navy and gold color scheme, with subtle texture suggesting hammered metal at the edges of each shape --tile --style raw --s 30 --ar 1:1 --q 2

Variation 3 — Abstract Watercolor:

/imagine A seamless watercolor wash pattern with loose organic shapes suggesting iris flowers and trailing vines in dusty mauve, sage green, and cream on a warm white ground. Soft edges, no hard outlines, painterly and imprecise --tile --s 100 --ar 1:1

Variation 4 — Dark Damask:

/imagine A seamless dark damask wallpaper pattern with intricate floral and scroll motifs, rendered in deep charcoal and black on a very dark grey ground with a slight satin sheen. Victorian gothic, formal and imposing --tile --style raw --s 80 --ar 1:1 --q 2

Video Generation

Ambient Loop:

/imagine A single candle flame burning in a dark room, ultra close-up, the flame swaying gently, light dancing on the surface of a wooden candlestick. --motion low --loop --ar 9:16

Dynamic Landscape:

/imagine A storm front approaching over flat prairie, anvil-shaped cumulonimbus towering above, lightning visible at the base of the cloud, golden wheat fields in the foreground bending to an increasing wind. --motion high --ar 16:9

Character Reveal:

/imagine A knight in ancient battered armor slowly turns to face the camera against a misty forest background. --motion low --ar 3:4

Abstract Flow:

/imagine Iridescent oil paint being dropped onto water, spreading and folding into complex fractal patterns. Deep teal, violet, and gold. --motion low --loop --ar 1:1

18. Prompt Building Workshop

Let's build a complex prompt from scratch, step by step. This walks through the thinking process that experienced V7 users apply.

The Brief

We want to create a piece for a sci-fi novel cover: a solitary space explorer discovers a derelict alien vessel. We want it epic, painterly, with a strong sense of scale and mystery.

Step 1: Define Your Subject and Lead With It

The subject is the explorer — a human character. That's our anchor.

A lone human explorer in a white exploration suit

Good start. But V7 rewards specificity:

A lone human explorer in a worn white EVA suit, helmet visor reflecting the alien ship ahead

Better. Now we have texture (worn), a specific detail (reflecting visor), and a directional relationship to the scene.

Step 2: Build the Environment

Where are they? What's around them?

standing at the edge of a massive circular airlock opening into the dark interior of a derelict alien ship

Let's add scale — this is important for that "epic" brief:

standing at the edge of a massive circular airlock — the explorer dwarfed by the sheer scale of the opening — looking into the dark interior of a derelict alien vessel

The phrase "dwarfed by the sheer scale" does visual work that no parameter can do.

Step 3: Add Atmosphere and Lighting

Lighting is the single most powerful tool. What kind of scene is this?

A cold blue starfield visible through the shattered hull sections, emergency bioluminescent strips still faintly glowing green along the walls

We've got contrast: cold deep space (blue) vs. decaying biological tech (green). That color opposition will create visual drama.

Step 4: Add the Style/Mood

What visual language should this evoke?

Painted in the tradition of Golden Age science fiction cover art — Chris Foss meets Moebius. Rich painterly detail, a slightly warm atmosphere in the ship's interior contrasting with the cold blues of space.

Specific artist references give V7 real aesthetic direction. "Cinematic" or "epic" alone are too vague; "Chris Foss meets Moebius" gives the model a coherent visual tradition to work within.

Step 5: Add Technical Specifications

Camera language still applies even in illustration:

Wide angle perspective exaggerating the scale, low horizon line putting most of the image above the explorer, deep detail in both foreground figure and distant interior.

Step 6: Assemble and Add Parameters

Full assembled prompt:

/imagine A lone human explorer in a worn white EVA suit, helmet visor reflecting the alien ship ahead, standing at the edge of a massive circular airlock — the explorer dwarfed by the sheer scale of the opening — looking into the dark interior of a derelict alien vessel. A cold blue starfield is visible through shattered hull sections. Bioluminescent emergency strips still glow faintly green along the organic-looking walls. Painted in the tradition of Golden Age science fiction cover art — Chris Foss meets Moebius — rich painterly detail, warm amber light spilling from deep inside the ship contrasting with the cold blue of space. Wide angle perspective emphasizing scale, low horizon line. --ar 2:3 --s 400 --exp 15 --q 2

Step 7: Iterate

First generation looks great but the explorer is too small. Let's use multi-prompt weighting to emphasize the figure:

/imagine lone human explorer in worn EVA suit::2 derelict alien vessel airlock, bioluminescent green interior, cold blue starfield::1, painterly science fiction, Chris Foss meets Moebius, scale and mystery --ar 2:3 --s 400 --exp 15 --q 2

Want more variation? Add --c 30 to the mix. Want a different feel on the style? Add --sref [relevant concept art URL] --sw 150.

Step 8: Lock It In With a Seed

Found a composition you love? Note the seed from the image metadata and add --seed [number] to generate variations that start from the same noise, giving you more control over the final polish.

19. Common Mistakes → Better Version

These are real V7 failure patterns with rewrites that actually work.

Mistake 1: The Quality Tag Pile-Up

Before:

/imagine beautiful woman portrait, stunning, gorgeous, 8k ultra hd, hyperrealistic, masterpiece, award-winning photography, perfect lighting, highly detailed, intricate

After:

/imagine A woman in her late 30s with strong features and kind eyes, photographed in a quiet moment of reflection. Soft natural light from a north-facing window, catching the silver threads starting to appear in her dark hair. Intimate, close-cropped, documentary portraiture. --ar 3:4 --style raw --s 40 --q 2

Why it works: Every word does specific visual work. No word is a generic quality signal.

Mistake 2: Style Before Subject

Before:

/imagine cinematic, dramatic lighting, moody, dark atmosphere, shallow depth of field, bokeh, a man standing in rain

After:

/imagine A man in a soaked grey overcoat stands at a bus stop in the rain, hands in pockets, looking down at his reflection in a puddle. The streetlight above halos through the rain. Cinematic, shallow depth of field, moody. --ar 2:3 --style raw --s 60

Why it works: Subject leads. Style descriptors are placed after the scene is established.

Mistake 3: Contradictory Instructions

Before:

/imagine a dark, moody, atmospheric scene that is bright, cheerful, well-lit, vibrant, energetic, and also melancholy

After:

/imagine A late-afternoon party starting to wind down. Most guests have left, leaving behind half-empty glasses and soft music still playing. One woman sits alone at a table by the window, the warm golden light catching dust motes in the air. Bittersweet atmosphere, gentle, contemplative. --ar 16:9 --s 150

Why it works: The "melancholy despite the warmth" mood is achieved through scene context, not contradictory adjectives.

Mistake 4: Using --cref (V6 Syntax)

Before:

/imagine Elena the warrior --cref https://i.imgur.com/example.jpg --ar 3:4

After:

/imagine Elena, a determined warrior in battle-worn silver armor, standing in a medieval great hall by firelight, hand on her sword hilt. --oref https://i.imgur.com/example.jpg --ow 120 --ar 3:4 --s 150

Why it works: --oref is V7's updated character reference system, replacing --cref entirely.

Mistake 5: Drowning the Subject in Context

Before:

/imagine beautiful magical ancient enchanted forest with glowing trees, mist, magical creatures, fairies, sprites, sparkling lights, mystical atmosphere, fantasy world, ancient ruins, waterfalls, colorful flowers, mushrooms, and in it there is a woman

After:

/imagine A woman in a hooded cloak pauses at the base of a massive ancient tree whose roots form a natural archway. The forest around her glows with soft bioluminescence — mushrooms, lichen, and trailing moss all faintly lit. Mist drifts between the trees. She looks up, as if hearing something. Fantasy illustration, mysterious. --ar 2:3 --s 300 --exp 10

Why it works: The woman is the subject; the magical forest is the context. The scene has a narrative moment ("as if hearing something") that organizes everything.

Mistake 6: Too Short for the Complexity Required

Before:

/imagine epic space battle

After:

/imagine Two capital-class starships engaged in close-range combat, their hulls lit by turbolaser fire and secondary explosions across their surfaces. In the foreground, a fighter pilot ejects as their small craft comes apart. The gas giant below provides an enormous amber backdrop. Epic science fiction concept art, massive scale, cinematic chaos. --ar 21:9 --s 350 --exp 15 --c 20

Why it works: A two-word prompt leaves everything to interpretation. Specificity in subject, environment, and key dramatic details produces a coherent scene.

Mistake 7: Not Using --no for Consistent Exclusions

Before:

/imagine A serene mountain lake at dawn, photorealistic

(Keeps getting hikers, boats, buildings)

After:

/imagine A serene mountain lake at dawn, the surface perfectly still, reflecting the first pink light on the peaks above. Dense conifer forest lines the shore. No human presence. --ar 16:9 --style raw --s 50 --no people, boats, buildings, dock, trail, campers --q 2

Why it works: --no explicitly excludes persistent unwanted elements.

Mistake 8: Expecting Natural Language to Work Without Specificity

Before:

/imagine A really nice flower in a beautiful vase with nice light

After:

/imagine A single white peony in a tall celadon ceramic vase, three petals fallen to the linen tablecloth below. Late afternoon light from a nearby window casts long soft shadows. Still life photography, Flemish painting tradition, intimate scale. --ar 4:5 --style raw --s 100 --q 2

Why it works: "Nice" and "beautiful" communicate nothing. Specific flower type, specific vase, specific fallen petals, specific light direction — each detail does real visual work.

Mistake 9: High --exp Destroying Prompt Accuracy

Before:

/imagine A young woman reading in an armchair by a window --exp 80 --s 300

(Result: dramatic, saturated, but barely resembles the prompt)

After:

/imagine A young woman reading in an armchair by a window, afternoon light across the pages, absorbed and still. Soft, intimate, warm interior light. --exp 15 --s 200 --style raw

Why it works: --exp 80 overwhelms the prompt's content. --exp 15 adds richness without hijacking the scene.

Mistake 10: Ignoring Aspect Ratio

Before:

/imagine A panoramic mountain landscape --ar 1:1

(Gets a cramped, awkwardly cropped view)

After:

/imagine A panoramic mountain landscape at golden hour, a glacier-carved valley below, distant peaks catching alpenglow, wispy clouds at mid-height. Wide-angle landscape photography, epic scale. --ar 21:9 --style raw --s 80 --q 2

Why it works: Aspect ratio should match the visual concept. Panoramic subjects need wide ratios.

20. Common Pitfalls

21. Cheat Sheet

Your at-a-glance reference for everything V7.

Core Parameters Quick Reference

Reference Parameters Quick Reference

Speed Parameters Quick Reference

Video Parameters Quick Reference

Model Parameters Quick Reference

Stylize (--s) Values at a Glance

Chaos (--c) Values at a Glance

Exp (--exp) Values at a Glance

Omni Reference Weight (--ow) at a Glance

Style Reference Weight (--sw) at a Glance

The V7 Prompt Template (Printable)

[WHO/WHAT] [specific details],
[WHERE/WHEN] [environmental context],
[MOOD/ATMOSPHERE],
[STYLE/TECHNIQUE],
[TECHNICAL CAMERA/LENS if applicable]
--ar [ratio] --s [0-1000] --exp [0-25] --style raw (if photorealistic) --q [1 or 2]

Example:

A retired astronaut [WHO] with cropped silver hair and laugh lines [specific details],
sitting at a kitchen table with a cup of coffee, looking out a window at a rainy street [WHERE/WHEN],
contemplative, bittersweet, the weight of years of adventure now quiet [MOOD],
intimate documentary photography [STYLE],
50mm lens, natural window light [TECHNICAL]
--ar 3:4 --style raw --s 50 --q 2

Quality Keywords: Use vs. Avoid in V7

Guide compiled and expanded from: Midjourney Docs, Blake Crosley Complete Guide, Printify Prompts Guide, SurePrompts V7 Guide, DataCamp V7 Tutorial. Last updated: April 2026.