Prompts and Circumstance
When it comes to generative AI tools like Midjourney, the quality of your prompts help determine the quality of the images you’ll generate. But what actually makes for a high-quality prompt? If you scour the internet you’ll find all sorts of advice for which prompts to use to get the best output. I’ve seen artists litter their prompts with long strings of redundant key words that supposedly make the difference between a lack-luster image and a great one.
In this article we’ll take a look at some of these key words to try to determine which ones actually make a difference, and which ones are just being ignored by Midjourney.
Disclaimer:
Midjourney is complex and is always evolving. What I suggest here today may work fine in version 4, but may be irrelevant by the time version 5 or 6 are released.
Moving Target
Using the prompt “a furry kitten” in each version of Midjourney shows how much more capable the tool has become in just a few months.
As I mentioned above, Midjourney is a tool that is being updated regularly. The images that are being produced with V4 weren’t possible with earlier versions. Therefore, it stands to reason that the hoops that artists had to jump through to get decent output from earlier versions of MJ are no longer needed in V4.
8k, 64k, UHD —Oh my!
I’ve often seen artists use keywords such as “8k”, “16k”, or “UHD” to ensure that it will result in more finely-detailed images. Does adding these keywords actually result in better results? Let’s find out.
As you can see, including the keywords “8k” or “16k” has little to no effect.
Highly detailed vs. highly-detailed vs. hyper detailed
Rather than improving the detail and fidelity, adding the keywords “highly detailed” or “hyper detailed” appears to produce strange artifacts that most artists would probably want to avoid. Adding a hyphen between the words seems to lessen the artifacts, but it doesn’t make a dent in improving the detail quality.
So if these keywords don’t have an effect, what does?
Distance from subject
If there’s one sure fire way to get better details from your MJ images it’s getting in close to your subject. Using keywords like “close up shot” or “extreme close up shot” will result in a higher level of detail, while having you subject further away will result in fuzzier details.
Adding the words “macro photography” or “macro lens” will yield similar results but with the added effect of a shallow depth of field:
Describe the details
Rather than littering your prompts with a lot of redundant and useless keywords about “quality”, try describing the details you’re looking to produce.
The two images below are upscaled from remixes of the same prompt “a line of furry kittens” but for the image on the right I added the phrase “soft delicate fur”. While there are still some issues with the kittens as the get further away from the camera, I feel like the overall results tend to be sharper with the added phrase.
Midjourney seems to like it when you embellish details with descriptive words.
What did we learned?
I think there are some clear takeaways here:
Using keywords like “UHD”, “8k”, “64k”, or “hyper detailed” to coax greater detail from Midjourney has zero positive effect, and may actually cause your images to look worse.
The sure-fire way to get the most crisp detail out of your images is to get as close as possible to your subject.
If your composition doesn’t lend itself to being framed close up, then using keywords that help describe details helps MJ in producing the results you’re looking for.
What techniques have you used to get MJ to render fine details and better fidelity?
If you have tips of your own, please share them in the comments below.
Also, if you like my articles, please share the link with others and let me know what other generative AI topics you’d like me to explore.