<100 subscribers

Midjourney offers you endless aesthetic control.
With basic text prompts, you get basic non-customized results here. For example, if you type "3 cats playing poker" into Midjourney without further tinkering, you might get something like these outputs:

Not awful. But if this is the limit of your prompting, you won't have any say in your outputs besides the base content of your scenes. If you want to direct your aesthetics more finely, you'll need to modify your prompts with Midjourney parameters.
Parameters are special shortcuts that instruct Midjourney to do certain things, and you can use multiple parameters together in one prompt to achieve better aesthetic control and more sophisticated outputs.
For instance, let's say you still want a "cats playing poker" scene, but you want your outputs to have a wide aspect ratio and to look more illustrated, surreal, and colorful. You could use a prompt like so:
a red demon cat, blue demon cat, and green demon cat sitting around a poker table casually hanging out --chaos 100 --ar 2:1 --exp 100 --sref 3002874870 --sw 500 --stylize 500
That scene prompt is more detailed than "cats playing poker," of course, but the ensuing parameters define the outputs' visuals. The parameters (which are always set off with two plain dashes) might look confusing at first, but the system is straightforward:
--ar tells Midjourney what aspect ratio you want, e.g. 1:1, 2:1, 16:9, etc.
--sref defines which Midjourney style reference code to apply; there are billions of these codes, and each one represents a unique aesthetic style (e.g. oil painting realism, ink wash minimalism, 3D claymation, and so on)
--sw is "style weight" and determines how strongly the style of your sref code should be applied; this parameter can be set to any number between 0 and 1,000
--stylize, which can also be set anywhere between 0 and 1,000, tells Midjourney how strictly or freely to interpret your prompt; lower values keep your prompt followed literally, while higher values give Midjourney more creative freedom
--chaos configures how much variety you get in your outputs; for every prompt you input, Midjourney produces four images; this parameter, which can be set from 0 to 100, lets you determine how much visual variation you see among your four outputs; a higher value will make your images more aesthetically distinct from each other
--exp, another 0 to 100 parameter, is a newer experimental modifier that works like --stylize and can be used to make more detailed and creative outputs

Accordingly, using multiple parameters together with good scene prompts lets you granularly direct the aesthetics of your outputs per your creative vision. You can create "style recipes" this way, too, that you can use to create consistent imagery over time.
With the example prompt that I broke down above, I created the featured graphic of this article. Notice how different the image is visually compared to the "3 cats playing poker" default-setting outputs I showed you first. This image is more expressive, more in line with my personal tastes:

Yet you don't have to memorize many Midjourney parameters to take better control here.
When it comes to aesthetic styling (color palette, texture, etc.), the bulk of the work stems from what --sref code or codes you use. I mention the plural "codes" because you can combine different sref codes to mix their aesthetic styles.
For example, in my example prompt above, I only used one sref code: --sref 3002874870. This code is responsible for the airy illustrative look and stark colors in this post's featured graphic.
If you used this sref code again in another prompt, e.g. "a shining Ethereum logo," and you did so without any other parameters applied, you'll still get a visual that's aesthetically in the ballpark of my featured graphic, like so:

So let's say you wanted to keep some of the aesthetic characteristics that --sref 3002874870 offers, but you wanted to additionally target a more 3D look.
Toward that end, first find another sref code that targets a 3D aesthetic, like --sref 275441216, which makes outputs like this:

Then to mix your target styles, you simply combine your sref codes (with a single space in between your codes) in your Midjourney prompts. For instance, this instruction...
a shining Ethereum logo --sref 3002874870 275441216
...tells Midjourney to make outputs like these:

As you can see, this style recipe gives you more of a 3D look while still orbiting the original color palette of --sref 3002874870.
Notably, you can also customize the strengths of your srefs with double colon syntax. In other words, if you write "--sref 3002874870::3 275441216" in a prompt, the "::3" tells Midjourney you want the first sref code to have triple the influence on your outputs compared to the second sref code.
Keep in mind that you can also use style codes and other parameters in tandem with reference image uploads to modify your prompts.
If you click on the image icon in the prompting box of the Midjourney web editor, you can upload references. Upload pictures to "Image Prompts" if you want to emulate their content, e.g. apples beneath a tree. Upload pictures to "Style Reference" if you want to emulate their aesthetic style, e.g. cubist portraits.

For example, I've noticed some sref codes struggle to make Ethereum imagery. To counter this, I can download a simple Ethereum logo from the web and then upload it into the "Image Prompts" section of my prompts to give Midjourney more inspiration.
Once you've produced some outputs, the Midjourney web editor offers you lots of flexibility if you want to keep experimenting. If you click on an individual output, you'll arrive in a dashboard filled with "Creator Actions" features.

You'll find these creator functionalities on the right side of an output's dashboard grouped by categories like "Vary," "Upscale," and "Remix." Personally, my favorite actions here are as follows:
Vary Subtle grants you four new outputs that will be only slight remixes of your target output
Vary Strong gives you four new outputs that will be very liberal remixes
Upscale Subtle produces a high-quality upscaled version of your desired piece.
Rerun offers four new outputs using the exact same prompt that produced your target output
Edit brings up a dedicated editing UI where you can resize and "inpaint," or change only specific areas, of your outputs
The "Edit" resource is particularly useful because it's common in Midjourney to get an amazing output that is 95% perfect but still has a few weird spots you want to fix. Use the brush tool in this UI to paint over the areas you want rerolled, so to speak, and then you'll get new outputs where only those areas have been changed.
At this point, you know everything you need to know to dive into Midjourney on your own. Now, the challenge is to curate sref codes that you like and would want to use in your personal or professional work.
Fortunately, you've got multiple good options for this curation:
"--sref random" in a prompt (as opposed to a specific code number) will tell Midjourney to pick a random sref code for your outputs; this is a great way to discover new styles, though it may take a bit to land on an aesthetic you enjoy
Style Explorer is a newer resource from Midjourney that serves as a hub for discovering new sref codes via large thumbnails; it's awesome for quickly honing in on styles that suit your tastes
Midjourney enthusiasts also share great sref codes on X all the time; two of my favorite follows in this regard are @gizakdag and @sergeantsref, but there are many more you might like as well
Lastly, I'll add that once you start amassing sref codes and specific style recipes, you can use Midjourney's folders system to keep track of everything and make new outputs in organized fashion. It's much easier to do this than scrolling back through all your outputs and trying to remember what you created and when.
Anecdotally, I use Midjourney to make graphics for my Metaversal articles, and I share some of my favorite personal creations on Zora. It's also just fun: now you can cast visual spells with words. If you use good prompts + custom parameters, you can bring anything you can imagine to life.
There's lots of ground I didn't cover in this post, like how to use pre-V7 sref codes or how to animate your outputs. But feel free to let me know if you have any questions here, and I'll point you in the right direction. I recommend bookmarking the Midjourney docs, too.
For now, I'll leave you with five of my favorite sref codes in case you want to jump into trying them yourself without hunting around first...





All that said, if you make some outputs after reading this, I'd love to see them! Tag me on Farcaster.
Share Dialog
1 comment
my X article from today now on @paragraph a beginner's guide to better aesthetic control in Midjourney V7 https://paragraph.com/@wmp/midjourney-style-primer-peaster