To use the Midjourney Describe feature, simply type “/describe” (without the quotes)followed by a space and attach an image.
The bot will generate a description of the image using natural language, so you can recreate a similar image.
This is a very simple process, but it can be confusing if you haven't walked through it before!
How to use /describe in Midjourney
Make sure you have the image handy that you want to describe.
Log into Discord and navigate to your own Midjourney server.
Type /describe and hit enter.
Click on the spot where it says to upload the image.
Grab the image from your computer, and upload it.
Hit enter (sometimes you have to hit enter twice).
The Midjourney bot will get to work!
It will respond with 4 descriptions of the image.
You can click the “imagine all” button to get a first rendition. But, if you're really watching your Midjourney hours (or have ridiculously slow internet and are on a time crunch), you'll want to carefully look at the words in each option and choose the one you like the best.
If you choose “imagine all,” you'll get all 4 created right away.
If you choose one of the options (I chose #4), a box will pop up and this is where you get to edit your prompts. I like this because often times, I'll put in an image and I want it to be a different aspect ratio.
After you've checked the prompt and added or taken away from it to make it to your liking, tell Midjourney to make the image by hitting the “submit” button! 🙂
Once the images are created, if it isn't what you had in mind, go back to the other options and see if there are some keywords that you can bring over into your prompt.
Sometimes this takes a little tweaking, depending on how intricate the image is, but I've had great results and never had to do more than 1 or 2 renditions.
You can use the reroll feature (the little button that looks like a recycle symbol) – but this can only be used when you see all 4 images at once (before upscaling) so scroll back up if you want to use it. You can see the re-roll button in the image I shared that has the 4 Halloween options.
You can use the vary subtle or vary strong feature to get new versions of the images you already created.
This is only an option after you have upscaled and there's one image showing.
This doesn't change them a whole lot, at least in my experience on version 5.2, but it can be fun to try. 🙂
You can also use the zoom out feature at this time. This will zoom out from your image – so if it's cut off, sometimes this can help remedy that issue.
If you have any questions about using the /describe feature, please leave them in a comment below or ask in our The Women of AI Facebook group – all women are welcome! 🙂