I’m going to show you how to turn AI into photoshop on steroids…
WAYYY easier than photoshop, even easier than using Canva!
First some basics.
For this “tutorial” you’re going to need the Midjourney discord server, Google it, it’s free to access.
Once inside you’re going to go to any of the /newbie channels and you’ll get to create your first 25 images for free.
To make an image first you need to type /imagine and it’ll give you a “prompt” screen. Now the fun begins.After typing /imagine you’ll give it a short sentence about what you want to see.
- Incredible Hulk arm wrestling Superman
- Mountain landscape with a log cabin and a river
- Statue of Liberty holding an iphone
Whatever your brainz desires to see.
Now for the advanced stuff…
If all you do is just give it a sentence, you’re not going to have much control over the output.
You need to understand “switches” which help define your prompt better to the underlying AI.
If you type -- after describing your image, you can get finely tuned controls. These -- commands are called switches.
First let’s talk about switches for getting the right image layout.The default, when you use no switch, is 1:1 (square).
Try these switches for different layouts…
--ar 2:3 = portrait mode
--ar 3:2 = landscape mode
Here’s some more “switches”
--hd will make your images come out in high definition
--v 4 will make sure you are using version 4 the latest version, MUCH BETTER IMAGES!
--test lets you use a test version which is sometimes better.
--testp for photorealistic versions of your image description.
--wallpaper automatically gives all the right settings to make digital desktop wallpapers
--niji will give you really cool anime images
Ok now talk negative modifiers…
Using these switches will help you to tell the AI what you don’t want to see in your images.
--no tells Midjourney what you don’t want in your image.
Example --no grass or --no animals
--uplight will give you less change when you upscale for the final version
More on “upscaling” later.
Another incredibly important switch you need to know about are seeds. Seeds are kind of like fingerprints.
If you use the same seeds for multiple images you’ll get way less change. Seeds help you refine one image instead of get whole new images.
If you want steady results as you improve your image you need to define a “seed.”--seed [insert number] is how you do this. Adding a number sets the amount of “noise” in the image and makes for much steadier results.
--sameseed sets the same seed for all 4 image variations
Now let’s discuss granular controls…
--s [number] sets how strict you want the AI to be with your description. Default is --s 2500, limit is --s 20,000.
Higher the number the more creative it will be. Lower the number and the more strictly it’ll follow your words.
--q 0.25 for fast low quality images (great for quick ideation)
--q 1 is the default
--q 2 means double the quality or 2 GPU minutes of quality upgrading
--q 5 is the limit but be careful as it can go off the rails
Ok now let’s put it all together…
I want to make a high quality image of a mountain scenery and photo realistic. See the picture below with the blue sky and photorealistic look, that was made with the prompt in the caption.
Now look for the image with the red ish brown grass. That image was made with the prompt and switches in the caption. This time using version 4, the latest greatest version of Midjourney’s AI.
Ok we like that one, but I want green grass and blue water in the river. AI isn’t a mind reader, it’s an assistant. Just like a real assistant you need to give it direction.
As you see in the image caption, you can add descriptive terms separated by comma in your prompts to add more details to your image. It’s AMAZING!
Try making that with photoshop, good luck.
You’d need to have MAJOR skills.
We did it by writing a sentence. Which kind of reminds of Genesis 1, in the Bible, when God speaks and forms mountains.
Sometimes the first image isn’t what you want. Click the 🔄 to have all new variations made.Keep doing variations until you get one you like.
When you find the one you like, you need to upscale it by clicking...
u1 = top left
u2 = top right
u3 = bottom left
u4 = bottom right
Upscale means it will finalize the image with increased detail in a higher quality image size. If you liked this tutorial, please give it a share on Twitter or Facebook.
It took me over 20hrs to learn all of this and you got it in under 20mins, hopefully that’s worth a share.