Midjourney advanced commands: a complete guide to mastering AI

La Artificial Intelligence Generative technology has completely changed the way images are created.Now, a simple text can be transformed into a stunning illustration, a hyperrealistic film still, or a professional-level concept art piece. Midjourney is one of the tools that has made the biggest impact in this area, and once you start mastering its features... advanced commands and Tricks prompt engineering, ceases to be “a IA that he draws” to become almost a creative superpower.

If you want to go beyond the typical “/imagine a beautiful landscape” To truly get the most out of Midjourney, you need to understand how its prompts, parameters, special modes, and the little hacks used by those who push it to its limits daily work. In this guide, I'll break down, step by step, everything covered by the top-ranked websites on "Midjourney advanced commands," but explained in clear, direct, and highly practical language so you can apply it to your own projects: from e-commerce campaigns to personal illustrations.

What is Midjourney and why is it worth learning its advanced commands?

Midjourney is an AI-powered image generator that interprets text descriptions. (prompts) and transforms them into illustrations, renders, photorealistic scenes, or concept art with spectacular aesthetic quality. It was originally launched exclusively within Discord, but now also has a web version, making it much more accessible to anyone who wants to create visual content without being a designer.

Unlike other models such as DALL·E or Stable DiffusionMidjourney is highly focused on producing images with an artistic, cinematic feel and rich in detail. It's ideal for marketing campaigns, mockups, covers, concept art, social media posts or blog postsFurthermore, with each new version (especially V5 and later) the anatomy, scene consistency, and even the dreaded hand generation have been improved.

The basic flow is simple: you type the command /imagine followed by your description, and after a few seconds, Midjourney returns a grid of four options. From there, you can improve one, request variations, or redo the whole thing. But where the possibilities really open up is when you start playing with parameters like –ar, –stylize, –seed, –chaos, –tile, –repeat and with techniques such as multiprompts, weights, or permutations.

Mastering advanced commands is not a technical whim.This is what makes the difference between "acceptable" images and results that look like professional commissions tailored to your brand or project.

How to structure a good prompt in Midjourney

A prompt in Midjourney can be as simple or as complex as you want.However, it's important to follow a logical structure so that the AI ​​understands what it needs to generate. This doesn't mean writing endless paragraphs, but rather clearly covering the key elements.

The basic structure of a complete prompt usually includes:

• Command: normally /imagine, which is what tells the bot to generate an image.
• Reference images (optional)Image URLs that will guide the style, composition, or character.
• Textual description: subject, action, environment, lighting, framing and key details.
• Style modifiers: artistic techniques, artists, genres, general aesthetics.
• Technical parameters: aspect ratio, level of chaos, quality, stylization, seeds, etc.

A well-assembled prompt might look like this. (generic example): /imagine heroic warrior on a snowy battlefield, dramatic lighting, cinematic, ultra detailed, oil painting --ar 2:3 --stylize 300 --chaos 10You don't need to recite a novel; the important thing is to cover the what, the how it looks, and the final format you want.

Another key idea is that Midjourney understands English better than Spanish.Especially when we're talking about styles, photographic techniques, or the names of artistic movements. You can write the idea in Spanish and then translate it into English with a reliable translator to get the most out of it.

Essential Midjourney Commands on Discord

Although the web version greatly simplifies the experienceIn Discord you have a series of commands (similar to custom slashes in SlackThese are the core of advanced work with Midjourney. They are not complicated, but each one opens a different door.

The most important ones for working seriously with advanced prompts are:

• /imagination: the main command to generate images from text and/or images.
• /blend: designed to quickly mix between two and five images, very useful on mobile.
• /describeYou send it an image and it returns four text prompts that describe it.
• / settings: opens a panel to configure model version, default style, Remix Mode, etc.
• /preferred option: allows you to create custom parameters that combine several parameters at once.
• /prefer suffix: automatically adds a text “tail” or parameters to all your prompts.

/blend is perfect when you want to merge visual references without complicating things by writing URLs one by one. You write /blendYou upload the images and let the model create a hybrid: ideal for characters that mix features, logos turned into illustrations, or creative mashups.

/describe is the star tool for stealing ideas from your own images For external resources: you upload the photo and Midjourney returns prompts full of styles, adjectives, and enhancers that you can then modify and reuse. It doesn't always nail the composition, but it's pure gold for finding styles and keywords you might not have thought of.

Midjourney versions, models, and special modes

Midjourney is not a single static modelbut rather a set of versions and modes that you can select depending on your needs. This directly affects which parameters you can use and the aesthetics of the result.

The elements to consider are:

• Model version: it is chosen with --v (1, 2, 3, 4, 5, etc.). It is normal to always use the most recent version.
• Style within each version: With --style You can change variants (for example, 5a, 5b, or the mode --style raw for something less “embellished”).
• Niji model: With --niji You activate a model specializing in anime and manga aesthetics.
• Test models: wearing --test o --testp You gain access to experimental variants that ask for feedback.

Changing the version or style can completely alter how the same prompt responds.A text that in V5 produces an almost photographic portrait, in a Niji model becomes an anime character with enormous eyes and flat colors. That's why it's important to always indicate the version and, if necessary, the style.

Also note that some parameters are limited depending on the version.For example, certain extreme aspect ratios or combinations of quality and size may not be available or may produce strange results on older models, whereas V5 supports more flexible aspect ratios and better detail with fewer artifacts.

Upscale, variations and Remix Mode: how to iterate until you nail the image

When Midjourney ends a generationIt returns a 2x2 grid of images. Below you'll see buttons U1, U2, U3, U4 (Upscale) and V1, V2, V3, V4 (Variations), plus some extra actions depending on the version.

In versions prior to 5Upscaling used to mean regenerating that image at a higher resolution, with more detail and, sometimes, slight differences in textures and shapes. In V5, images are already generated at 1024x1024 by default, so the U buttons simply "detach" that image from the grid to work with it independently.

The variations (V1-V4) generate four new images Based on one of the previous versions, maintaining the same prompt and a similar initial noise image. This allows you to fine-tune pose, lighting, or composition without completely losing the original idea; it's like asking an illustrator, "Make me four more versions of this concept."

Remix Mode is the icing on the cake for iterating: if you activate it in /settingsEach time you press a variation button, a small window will open allowing you to edit the prompt before generating the new images. This way, you can change characters, backgrounds, styles, or certain parameters while maintaining the original visual structure.

However, Remix doesn't start from scratch.It takes into account the generated image, its composition, and its elements, and "transfers" them to the new prompt. Subtle changes (adding a weapon, changing the armor type, altering the environment) usually work very well; drastic changes (going from a warrior to a coffee cup) produce strange but sometimes interesting results. You have to experiment and see how far the consistency holds.

Advanced parameters in Midjourney: the true dashboard

Parameters are options that you add to the end of the prompt. to control how the image is generated at a technical and aesthetic level. They always use two hyphens (--) and you can combine as many as you want as long as they are compatible with the version and model you are using.

In Midjourney version 5, the key parameters you should know are::

Aspect ratio: –aspect / –ar

Parameter --aspect o --ar defines the width:height ratio of the imageBy default, Midjourney generates a square (1:1), but you can change it to more photographic or vertical formats for characters, covers, banners, etc.

Typical examples of aspect usage would --ar 3:2 for a landscape, --ar 2:3 for a vertical poster or --ar 16:9 For a widescreen format. V5 accepts virtually all reasonable aspect ratios, although when you go beyond 2:1 or 1:2 it may start cropping poorly or compositing strangely.

It's not just a matter of final size.The aspect ratio influences the initial noise image and, therefore, the entire composition. The same prompt in 1:1 and 16:9 can radically change how the scene is organized and which elements appear cropped or centered.

Chaos: –chaos / –c

Parameter --chaos It controls how unpredictable the outcome will be. With --c 0 The model tries to be more stable and "obedient"; as you go up to 100, the AI ​​lets its hair down and allows itself weirder solutions, strange compositions, and visual experiments.

A high level of chaos is ideal for the exploration phase.: when you're still not quite sure what style you want, when you're looking for crazy ideas, or when you want to break up the monotony of the prompt. Once you find a line that works for you, you can reduce the chaos to achieve more consistent variations.

Image weight: –iw

When you use images as part of the prompt (image prompting), --iw It allows you to decide how much weight the visual element has over the text. The default value is 1, and you can usually adjust it between 0.5 and 2.

If you go up --iwMidjourney adheres more closely to the composition, form, and style of the reference image. If you lower the level, it relies more on the textual description and uses the image only as light inspiration. This is very useful for redesigning characters, changing their environment, or adapting a rough drawing into a decent illustration.

Negative prompt: –no

Parameter --no It's used to tell Midjourney what you DON'T want to appear.Everything you write after --no and before the next parameter, it will be taken as elements to avoid in the scene.

It is especially useful for removing annoying details: unusual text and fonts, extra fingers, repetitive background elements, colors that don't match your brand, etc. Typical examples would be --no text, letters, logo o --no hands When you want to recreate a pose using a previous image but without distorted hands.

It works better with concrete concepts than with very abstract things.And it's not foolproof. If you want finer control, you can combine it with negative multiprompts, which allow you to adjust the weight of what you want to "push out" of the image.

Quality of detail: –quality / –q

--quality Control how many resources Midjourney dedicates to that imageIn previous versions you could put --q .25, --q .5, --q 1 or even --q 2 to fine-tune how much detail you wanted. In V5, the maximum value is 1; you can't exceed it to get more detail, but you can decrease it to save time and credit.

Lowering the quality is perfect for quick sketchesThis is useful for composition tests or when you want to generate many inexpensive images to later select the best ones and redo only those with more care. Remember that quality is not resolution: resolution is determined by the model; here you adjust the fineness of the details.

Repetition: –repeat / –r

With --repeat o --r You tell Midjourney to execute the same prompt several times in a row. For example, --r 3 It will generate three separate works, each with its own grid of 4 images, from the same text.

It's a massive time saver when you're fine-tuning a prompt And you want to see many options at once. Furthermore, if you combine that with a reasonable level of chaos, you multiply the chances that "the gem" you were looking for will appear in one of those grids.

Seeds: –seed

The seed or --seed is the number that labels the initial noise image From this point, Midjourney begins to "sculpt" the final image. If you repeat the same prompt with the same seed in the same version, you will obtain virtually identical images.

Using seeds allows you to have relative control Regarding variation: you can experiment with different styles, aspect ratios, colors, or minor text changes while maintaining the overall compositional concept. This is very useful for producing coherent series or for correcting a specific detail without ruining everything else.

In V5 you can no longer retrieve the seed of a scaled image using the envelope emojiBecause in this version there is no second upscaling process: the images are already output at maximum resolution. But you can set the seed yourself with a number between 0 and 4.294.967.295 and reuse it when needed.

Stop: –stop

Parameter --stop stops the generation process prematurely, in a percentage that you choose between 10 and 100. With low values, the images become more "blurry", less detailed, as if they were a sketch or an unfinished render.

It's a rather niche parameter and doesn't always have a clear use.But it can be useful if you want a more painterly look, with less definition, or if you're simply interested in seeing how a scene evolves without going all the way to denoising.

Stylize: –stylize / –s

--stylize It is one of the most important parameters in Midjourneybecause it tells you how much weight to give to your artistic training in relation to your prompt. In V5 the default value is 100 and you can move it from 0 to 1000.

With a low stylize, the model clings more to the text And it gives you results that are perhaps less flashy but more faithful to the description. With high values, the AI ​​puts everything it has learned from user-voted images on the table: more detail, better composition, more drama… but also a greater tendency to ignore certain parts of the prompt.

It's not the same as chaosChaos focuses on "crazy" creativity and variation, while Stylize pushes towards prettier, more complex, and polished visuals. In short prompts, increasing Stylize makes little difference; in long prompts with many elements, it can radically transform the scene.

Mosaic patterns: –tile

Parameter --tile It generates images that can be repeated in a mosaic pattern. Seamless, perfect for textile patterns, wallpapers, or repeatable graphic resources.

The classic trick is to generate a simple tile. (for example, pastel-colored drops) and then use it as a base texture for costumes, backgrounds, or interfaces. Downloads You take the image, repeat it in an editor or on a website that checks for continuity, and if it fits well, you have your pattern ready.

Advanced prompting: starting images, multiprompts, weights and permutations

Beyond the technical parameters, the real leap forward comes when you understand how to structure the text and how to combine it with images to guide the model without overwhelming them.

Midjourney accepts both images and text as input.You can use them separately (only images with /blend or just text with /imagine) or mix them: one or more images at the beginning of the prompt and then a written description that refines what you want to achieve.

If you are going to use multiple images as a referenceIt's recommended that all the frames have a similar aspect ratio to the final image to avoid strange compositions. And remember to play with... --iw to adjust how much influence you want them to have on the text.

A commonly used technique is: first you mix images to obtain a base character (for example, by combining a classical sculpture, a series still, and an anime drawing), and then you use that image as a unique reference alongside an increasingly elaborate textual prompt to define action, environment, lighting, style, and detail.

Multiprompts with :: and weights

Although it may seem that commas separate ideas within the promptIn reality, Midjourney only pays limited attention to them. If you truly want to segment distinct concepts, you need to use the multiprompt separator. ::.

A multiprompt would be something like viking warrior with shield:: full body:: intense action scene:: oil painting:: highly detailedEach block separated by :: It is treated more independently, which helps prevent AI from confusing compound concepts (such as "cotton candy" or "hot dog") with free combinations of words.

Additionally, you can add a weight number after the last one. :: to indicate which part of the prompt should be given more relative importance. For example, cheese:: cake::1.5 This makes the cake more important than the cheese, compared to cheese cake, which tends to interpret “cheesecake” as a proper dessert.

The weights are relative to each other: if you put cheese::1.5 cake::1.5This is equivalent to not having used numbers at all, because both carry the same weight. The interesting thing is to play with 1:2 or 1:3 ratios to emphasize one concept without negating the others.

You can also use negative weights, type hands::-0.5, so that an element behaves like an “anti-prompt”, similar to --no but with intensity control. This usually works best for removing very specific things or for adjusting stubborn details that refuse to disappear.

Permutations with { }

Permutations are a spectacular tool for testing many variations. from the same prompt without having to rewrite it ten times. They are used with curly braces. { } and commas to separate options.

For example, this prompt {viking warrior with shield, muay thai fighter, reptilian}, full body, intense action scene, oil painting, highly detailed --ar 2:3 It will generate one job for each of the three subject options, keeping everything else the same.

You can nest permutations to vary specific details, like the weapon: {viking warrior with {sword, axe, shield}, muay thai fighter, reptilian}, full body...This allows you to generate huge test sets in a very short time, but be careful, because only the first 40 combinations will be processed and you can use up your fast processing time without realizing it.

It is also possible to permute parameters and parts of multiprompts: For example, viking warrior with shield, full body --ar {3:2, 2:3} o with shield::{2, 1.5}It's the quickest way to find out what aspect ratio, weight, or style combination works best for you.

If an option within the curly braces includes commas that you want to preserve As part of the same block, they must be escaped with a backslash. \ so that Midjourney doesn't treat them as permutation separators. This way you can permute "packs" of complex styles without them breaking down.

Other creative parameters: weird, tile, repeat, and the like

In addition to the parameters already discussedThere are a few extras that can help you explore rarer terrains or generate specific resources faster.

--weird o --w pushes the model towards strange interpretationssurreal or downright alien. It accepts values ​​between 0 and 3000: the higher the value, the more outlandish the AI ​​becomes. It's perfect for generating dreamlike images, cosmic horror, experimental art, or simply for breaking away from the typical.

--repeat We've already seen itBut combined with chaos or weirdness, it becomes a creative idea-spitting machine; ideal when you're in the visual brainstorming phase for a campaign or an artistic project.

--tile As already mentioned, it is key for repeatable patterns which you can then use in illustration, product design, interfaces, or backgrounds. Along with image blending and the parameter --iwIt offers a lot of possibilities for creating "fictional" textiles, wallpapers or coverings that look like they've been taken from a real catalog.

Prompt engineering tricks for better control of results

When you start using Midjourney intensivelyYou realize that there are recurring patterns: elements that don't appear no matter how much you mention them, details that the AI ​​invents, styles that get out of hand... That's where small prompt engineering tricks come into play.

A useful strategy is to duplicate parts of the prompt in multiprompts. When something doesn't quite disappear or reappear, if a single negative weight isn't enough to remove it, you can add that part twice with balanced negative weights to reinforce the message without breaking the overall weight.

Another practical technique is to combine a full “normal” prompt with an extra multiprompt Finally, dedicate this to emphasizing a detail you're missing. For example, if your character doesn't have tattoos despite mentioning them, you can add :: tattoos::2 at the end of the prompt so that Midjourney puts more focus on that specific element.

And if they still ignore you, it's time to get a little rough and repeat the word several times.Writing "tattoos tattoos tattoos..." often convinces the model that it's absolutely important. It's not elegant, but it works surprisingly well in some rebellious cases.

Don't forget the command /describe as a vocabulary generatorYou can create crazy images using chaos and then run them through /describe so that the AI ​​itself suggests styles, authors, and technical terms that you can reuse in future prompts with more control.

Practical applications: ecommerce, social networks, blogs and internal resources

All this arsenal of commands and parameters is of little use if you don't apply it to real-world cases.Midjourney fits especially well in contexts where you need personalized, fast images with a defined aesthetic, but you don't have the budget for complex photo shoots.

In ecommerce you can use it to create campaigns that are impossible to photograph.For example, electric guitars that work underwater, products in extreme environments, or surreal scenes that reinforce brand attributes. Instead of using generic stock images, you design custom visuals that reflect your aesthetic.

For social media and blogs, Midjourney lets you illustrate every piece of content with original images: educational articles, inspirational posts, carousels, video covers, etc. You just need to adjust the aspect ratio to what each platform requires (for example, --ar 9:16 for stories or reels) and play with styles consistent with your visual identity.

AI-generated mockups are also very usefulModern offices, gallery walls with artwork, packaging on study tables, posters on marquees… You can create complete scenes where your product is imagined “in context”, without having to physically set up the scene.

Internally, Midjourney serves as a style laboratoryMood boards, color usage examples, presentation mockups, iconography, backgrounds… You can even create images to illustrate brand manuals, visual tone guides or sales presentations, maintaining an aesthetic consistency that would be very expensive to produce in a traditional way.

Ultimately, the key is to combine knowledge of advanced commands With a clear vision of what you want to convey visually: AI does the hard work of rendering, but the creative direction remains yours. The better you are at writing and structuring prompts, the more it will seem like you have an entire team of illustrators working for you, even though you're actually just chatting with a bot on Discord or the web.