Summary
If you’ve been dabbling in Midjourney, but have hit a wall or aren’t getting the results you want, there are a few simple things you can do to get more out of the software, and learn how flexible it can really be.
Use the Web Interface
At launch, the only way to use Midjourney was by way of the Discord chat app. You had to input commands to a Discord bot, which meant learning quite a lot of options and syntax. In the meantime, Midjourney has launched a web interface atMidjourney.comwhich afarmore user-friendly interface. If you’re still using the Discord interface, I’d strongly suggest switching over to the web interface, where you can concentrate on creating rather than remember which commands do what.
Get to Know the Tools
Once you’re on the web interface, you’ll see all the tools laid out for you. To get started, you may have a look at myMidjourney web interface guide, but this is a good time to simply look at all the knobs and buttons laid out for you. Experiment with each one, see how they affect your images, and how they can work in combination to produce more diverse results.
In particular, make use of the editing tool after you first generate an image. You can use this to fix errors in an otherwise good image. See a hand with the wrong number of fingers? Just paint it out and try again.
Pick the Right Model
There are various Midjourney models, with version 6 being the latest as of this writing. You can always change to one of the older models, though anything older than version 5 isn’t that useful. Still, some people feel the older models were better at some things, or that the prompting style worked better for them. Also, if you want to create anime-style art specifically, switch to theNiji model.
Prompt Clearly
It’s not just the image generation portion of Midjourney that’s been getting steadily better over time. The model’s understanding of prompts has also become much more advanced. The latest model has revamped prompting, and you can now use more natural language than before.
Longer, detailed, and specific prompts work best in my experience, though you will still get results with a short and general prompt. Don’t be afraid to write a long prompt. One trick I like to use is having ChatGPT compose the image prompt for me based on my description of what I’m looking for in the output.
I have found the big fleshed-out ChatGPT prompt almost always gets better results.
Learn Your Parameters
There are variousparameters and commandsthat can be executed in Midjourney, and it’s a good idea to take the time and learn them all. These can be as simple as adjusting the aspect ratio, or as complex as changing how stylized or strict your output is.
Rate the Community’s Work
If you look at the above image, you may notice a “Personalize” button in the bottom left corner of the parameters page or on the lefthand sidebar of the web interface. This lets Midjourney apply stylization to your images that are designed to appeal to your specific taste. How? Well, every time you rate images in the Midjourney image feed, your preferences are stores to build up a sense of your personal aesthetics.
Of course, you need to rate a minimum number of images for this feature to become available. The easiest way to do this is to click the “unlock” button next to personalize and rate images until you’re approved. The more you do this, the more effective your personalization will be. Sometimes we’re not entirely sure why we don’t like a picture, but Midjourney’s algorithm can pick up on your aesthetic preferences even if you can’t clearly aritculate them.
Midjourney has become significantly easier to use and create consistent results since its launch, and so, if you were intimidated before, now is the best time to become familiar with the service, and how far it’s images have come.
