The Complete Midjourney Prompt Guide

Midjourney is capable of results that rival professional photography and illustration, but most people who use it never get there - not because the model is limited, but because most prompts are closer to search queries than creative briefs. A prompt like "woman in a red dress" specifies one decision and leaves five or six others unaddressed, and Midjourney doesn't leave those blank. It fills them in with whatever is statistically safest across its training data, which is exactly why so much AI-generated imagery has that faintly generic, interchangeable quality - flat lighting, centered framing, a color grade that reads as "professional" but also reads as nothing in particular.

This guide exists to fix that at the root. Rather than treating Midjourney as a collection of unrelated tips and parameter tricks, everything here connects back to one underlying structure: the 6-layer prompt framework. Once that framework is genuinely understood, every other Midjourney topic - which parameter to reach for, how to keep a character consistent across a shoot, when to use remix mode versus vary region - becomes a specific, learnable refinement instead of a new thing to memorize from scratch.

We've published 18 in-depth guides covering every corner of Midjourney prompting, from absolute-beginner mistakes to advanced multi-prompt syntax. Rather than making you find them one at a time through search, this page is the single hub - the framework explained once, then a complete map of every guide, organized by what you're actually trying to do. Bookmark this page if you're serious about Midjourney; it's the fastest way back to any specific topic without re-reading the basics each time.

Every genuinely strong Midjourney prompt - whether it's a product shot, a portrait, or an abstract brand visual - specifies the same six things. You don't need to max out all six in every single prompt, but the more of them you address deliberately, the less room Midjourney has to default to its safest, most generic guess. This is the single idea every other guide in this hub builds on.

1. Content type. What kind of image is this - portrait, product still life, flat lay, editorial, abstract? This frames every decision that follows and should almost always be the first thing you specify.

2. Subject. Describe your subject the way a film director would - material, pose, expression, texture, context - not just the noun. "A woman" gives Midjourney almost nothing to work with; "a woman in a tailored ivory linen blazer, relaxed posture, soft confident expression" gives it a real brief.

3. Colors and materials. Hex codes mean nothing to a diffusion model. Describe color as material behavior in light - "deep burgundy velvet" instead of "red," "warm champagne satin" instead of "gold." This is also where most people default to nothing at all, which is a large part of why so many AI images share the same washed-out palette.

4. Composition. Tell Midjourney where things go in the frame - rule of thirds, generous negative space, centered versus off-center, full body versus close crop. Composition is direction, not decoration, and it's one of the fastest ways to break out of the model's default centered-symmetrical habit.

5. Lighting. Source, direction, and quality. Soft diffused window light from the left feels intimate; hard direct sun feels harsh; golden hour backlight feels nostalgic. Of all six layers, lighting is consistently the single highest-impact addition relative to the length of the phrase it takes to write.

6. Camera and lens. A specific focal length and aperture - "85mm lens at f/2.0" - tells Midjourney exactly how much background blur and perspective compression to apply. "Shot on 35mm film" signals grain, color rendering, and a specific imperfection level that reads as authentically photographic rather than synthetic.

Read that list again and notice what it doesn't include: a single parameter. Parameters (covered below) modify how Midjourney interprets a prompt; the six layers are what the prompt actually says. Get the six layers right first - parameters are a multiplier on a good prompt, not a substitute for one.

Three guides cover the foundational skill of writing (and rewriting) prompts using the 6-layer framework itself.

The full walkthrough of the 6-layer framework in practice - why most prompts fail (they specify only the subject and leave everything else to chance), how to build a prompt layer by layer, and a direct before-and-after comparison showing the same subject with and without the framework applied. Covers which Midjourney parameters matter most once your prompt structure is solid, plus the specific beginner mistakes that keep results looking flat. If you only read one linked guide from this hub, this is the one - everything else assumes you've internalized this first.

Read the full guide: How to Write Midjourney Prompts That Work →

Explains the specific mechanism behind the "AI look" - not a Midjourney flaw, but what happens whenever a layer of your prompt is left undefined and the model defaults to its statistically safest guess. Covers the research behind this behavior and the three invisible defaults (flat lighting, centered framing, a safe neutral color grade) that flatten most AI-generated images, with concrete language to override each one. This is the diagnostic companion to the writing guide above - read it when your prompts are structurally fine but the results still feel forgettable.

Read the full guide: Why Your AI Images Look Generic →

Negative prompts (Midjourney's --no parameter) tell the model what to actively exclude rather than just omitting it from the main description - the difference matters because leaving something unmentioned doesn't guarantee Midjourney won't generate it anyway. This guide covers the syntax, the most common practical uses (removing unwanted text, watermark artifacts, or a recurring stylistic problem), 30 ready-to-use negative-prompt phrases, and why --no works best as a targeted fix for a specific recurring issue rather than a blanket default added to every prompt.

Read the full guide: What Is a Negative Prompt in Midjourney →

Four guides cover the parameters worth actually learning, from a full reference down to the two most commonly misunderstood settings.

The full reference: every parameter worth knowing, what it does, and when to use it - --ar for aspect ratio, --stylize for creative interpretation strength, --chaos for output variation, --raw for disabling the default aesthetic filter, --seed for reproducibility, --v for model version, and --oref/--cref for reference-driven consistency. Written around real side-by-side examples rather than abstract definitions, so you can see what each setting actually changes instead of just what it's named. Start here if you've never deliberately set a parameter beyond aspect ratio.

Read the full guide: The Complete Guide to Midjourney Parameters →

A deep dive into the single most misunderstood Midjourney setting. --stylize controls how much of Midjourney's own aesthetic judgment gets layered on top of your literal description - low values follow your prompt closely and can look flat, high values look more polished but drift further from exactly what you wrote. This guide walks through real examples at every major value range and gives starting-point recommendations by content type (product photography, portraits, editorial, abstract work), built from direct side-by-side testing rather than guesswork.

Read the full guide: Midjourney Stylize Parameter Guide →

Covers the full --ar ratio range and, more usefully, which ratio actually fits which platform and shot type - 4:5 for Instagram feed, 9:16 for Reels and TikTok, 1:1 for versatile cross-posting, 16:9 for landscape and banner work. Also explains how aspect ratio changes composition itself, not just the crop - a scene built for 16:9 rarely just crops cleanly to 4:5, so setting the correct ratio at generation time produces a meaningfully better result than generating square and cropping after the fact.

Read the full guide: Midjourney Aspect Ratio Guide →

Midjourney V8.1 is the current default model, and this guide covers every meaningful change from V7 - 5x faster generation, native 2K resolution, dramatically improved text rendering, and better prompt understanding, all tested directly rather than summarized from release notes. If you've been using Midjourney for a while and haven't checked what changed recently, or you're new and want to know what you're actually working with, this is the version-specific reference to read alongside the general parameters guide.

Read the full guide: Midjourney V8.1 Complete Guide →

Three guides cover keeping a character, style, or aesthetic consistent across multiple generations - a common pain point once you move past one-off images into a real content workflow.

Everything on Midjourney's character-reference system (--cref/--oref in newer versions) - syntax, weight control for how strongly the reference influences the result, combining character reference with style reference, and its real limitations, since --cref is genuinely useful but not magic and breaks down in some specific, documented situations. Essential reading if you're generating the same person or brand mascot across multiple scenes and need them to actually look like the same subject each time.

Read the full guide: Midjourney --cref Complete Guide →

A companion to the --cref guide covering what the --seed parameter can and can't do. Seeds lock composition and overall style, not a character's face across different scenes - a distinction that trips up a lot of people expecting seed-based consistency to behave like character reference. This guide explains exactly where seed's usefulness ends and when you genuinely need --cref instead, so you're not fighting the wrong tool for the job.

Read the full guide: Midjourney Seeds & Consistency →

Midjourney's personalization feature learns your aesthetic preferences from a set of ranked image pairs and applies that learned taste automatically via the --p parameter. This guide walks through building a profile from scratch, when it's worth the setup time (mainly for brands and creators generating a high volume of images who want a consistent house style without re-specifying it every prompt), and how personalization interacts with the other layers and parameters you're already using.

Read the full guide: Midjourney Personalization Profile Setup →

Three guides cover Midjourney's less commonly used but genuinely powerful syntax for weighting, testing variations, and mid-generation edits.

Covers how Midjourney's double-colon weighting syntax works - assigning relative importance ratios between prompt segments, using negative weights to push away from an element, and decimal weight support, which varies by model version. Also covers combining weighting with permutation syntax for more controlled batch testing. Most useful once you've hit a specific problem the basic 6-layer framework alone doesn't solve - a prompt where one element keeps overpowering another regardless of word order.

Read the full guide: Midjourney Prompt Weighting Guide →

Curly-bracket permutation syntax lets you test multiple colors, materials, or styles in a single prompt submission instead of running separate generations for each variant - genuinely useful for rapid options testing on a product or campaign. This guide also covers the separate double-colon multi-prompt syntax, which splits emphasis between distinct prompt segments rather than testing variants, and clarifies exactly when to reach for each, since they're easy to confuse but solve different problems.

Read the full guide: Midjourney Permutations and Multi-Prompts →

Remix Mode still works, but its behavior and interface have changed enough that a lot of older tutorials are now inaccurate. This guide covers how Remix Mode functions in the current version, how it differs from Vary Region (a common point of confusion since both let you modify part of an existing generation), and when each is the right tool - Remix for changing the prompt behind an existing image, Vary Region for editing a specific area while keeping the rest untouched.

Read the full guide: Midjourney Remix Mode Explained →

Five guides cover practical decision-making - beginner mistakes, ready-to-use prompt libraries, product photography workflows, and how Midjourney compares to the other major AI generation tools.

The 10 most common mistakes people make when they're new to Midjourney - from skipping prompt layers entirely to ignoring aspect ratio to fighting the model with contradictory instructions - each paired with the exact fix. A useful diagnostic checklist if your results have been inconsistent and you're not sure why; most beginner problems trace back to just two or three of these ten, and this guide helps you identify which ones apply to your specific prompts.

Read the full guide: 10 Midjourney Beginner Mistakes →

30 ready-to-use prompts organized from easy to advanced, built specifically to demonstrate the 6-layer framework in action across different subjects - portraits, products, landscapes, abstract compositions. Copy one, swap the subject for your own, and you have a working example of what a fully-layered prompt looks like in practice, which is often the fastest way to internalize the framework compared to reading about it in the abstract.

Read the full guide: 30 Midjourney Prompts for Beginners →

40 ready-to-copy prompts across 8 named aesthetic styles - minimalist, cottagecore, dark academia, Y2K, and more - each built on the same 6-layer structure but tuned to a specific visual mood. Useful both as a direct prompt library and as a study reference for how the same framework produces very different results depending on which specific words you choose for colors, materials, and lighting within each layer.

Read the full guide: 40 Midjourney Aesthetic Prompts →

The best AI product photography workflow in 2026 combines two tools rather than relying on Midjourney alone: Midjourney generates the scene and mood, while Nano Banana composites your real product photo into that scene with accurate lighting, shadow, and perspective - preserving your actual product rather than generating a fictional one. This guide walks through the exact step-by-step method, useful for anyone selling physical products who wants styled photography without a full studio shoot.

Read the full guide: The Two-Step Product Shot Method →

A direct side-by-side comparison covering image quality, ease of use, pricing, and text/logo rendering (where ChatGPT's image generation has a real edge over Midjourney). Written for brand creators trying to decide between the two rather than as a general feature comparison - the honest answer is most serious users end up needing both, and this guide explains specifically which situations favor which tool.

Read the full guide: Midjourney vs ChatGPT Images →

Midjourney and Sora 2 aren't really competitors in the way they're often framed - one generates images, the other generates video, and most serious content workflows end up using both together (a Midjourney still as the visual foundation, animated or extended in Sora). This guide covers exactly when you need which tool and the practical workflow for using them in sequence rather than treating the choice as either/or.

Read the full guide: Midjourney vs Sora →

If you've read this far, you already understand more about deliberate Midjourney prompting than the vast majority of people using the tool - most never move past the subject-only prompt this guide opened with. The fastest way to put the 6-layer framework into practice without holding all six layers in your head every time is to use a structured tool built around it: our free Midjourney Prompt Builder walks you through each layer with live preview, and the AI Image Prompt Generator builds all six layers automatically from a plain-English description if you'd rather describe your idea and let the structure happen behind the scenes. If a prompt you've already written isn't producing the results you want, the Prompt Fixer diagnoses which specific layer is missing and rewrites it for you.

This hub will be updated as new Midjourney guides are published and as the underlying model changes - bookmark it as your entry point back into any of these 18 topics, rather than searching for each one individually.