top of page

Why I Love the New /Describe Feature in Midjourney for Prompt Engineering

With their new /describe prompt, Midjourney has given us a look behind the curtain on how they create the images we see, letting us become better prompt engineers in the process.






I'm told that other AI art tools have already rolled out featured like this, but I'm still super psyched about Midjourney's /describe feature, because it's awesome. Okay? Just let me be excited about it.


Here's what I did with it on my first try. I liked it so much, I literally added it to the main Midjourney page on my website.


All you have to do is upload an image into Midjourney and it will give you four prompts that could have been used to generate such an image.


It's a prompting paradise!


So here's what I put in: something I had made with Midjourney a little while ago. It's basically a funky guitar man wood guy. I'm not exactly sure what he is, but Midjourney is about to tell me!




I was already really happy with this image. It's fun, it could easily be used as some sort of stand-alone image, basis for a logo, whatever. But I wanted to see what Midjourney would do with this when I used the new /describe function.


Here's what it gave me:


"1) an odd sculpture on a table with colored striped design, in the style of brown and beige, digitally enhanced, presentation of human form, rendered in maya, inlay, mechanized forms, pensive poses

2) dana dano kunfle in sculpture - 3ds max - 5, in the style of ndebele-inspired motifs, earthy colors, figura serpentinata, inlay, robotic motifs, disfigured forms, horizontal stripes

3) 3d sculpt sculpture by matt taylor, in the style of ndebele-inspired motifs, digital constructivism, brown and beige, graceful poses, horizontal stripes, neo-mosaic, figura serpentinata

4) 3d rendering of a statue with a broken guitar, in the style of ndebele-inspired motifs, wood veneer mosaics, kay sage, horizontal stripes, disfigured forms, terracotta, ahmed morsi"


Um, YES! That's freaking awesome! I never, ever, ever would have come up with all of those terms on my own thinking through what I wanted for that image. EVER. Not ever. Whoa!


Okay, so now the really fun part.


You put any one (or all) of those prompts RIGHT BACK into Midjourney and then play around with the results. Right away I loved some of the results it gave me.






All different interpretations, all awesome.


Then, I blended and did a very small amount of upscaling and varying utilizing this other image I had generated with Midjourney. I decided I wanted her to be a ballerina after one of the images gave that effect. It only took about 3 iterations and I ended up somewhere I really would not have expected, with an image I really loved. After a tiny bit of editing with my new favorite toy, PhotoRoom (not sponsored), I was done!


And that is the beauty of this new /describe prompt!























Written by the human staff @ Prompt Freely.


Comments


Commenting has been turned off.
bottom of page