Midjourney is one of the most popular AI image generators, and you can use it to create incredible artwork in just seconds. In this video, I'll show you how to get started, craft the perfect prompts, and unlock advanced options for your dream designs. Let's dive in.
Now to access Midjourney, go to the following website. You can click on the card in the top right or the link in the description below. Once you're here, click on the signup link in the bottom right-hand corner.
Now I know that all you cool kids want to continue with Discord, but we're going to cover the Midjourney web experience today, and the fastest way to get started is to go with Google. So, click on the continue with Google button, type in your Google account credentials, and I'll see you on the other side. Okay, so you've signed up successfully, and now you've landed on the explore tab.
This tab is here to inspire you with the possibilities of what you can create with Midjourney. You can click on any image to see the prompt and options used to create that image. On the right side of the screen, you can see the prompt and the options are listed below.
We'll cover all these in a minute, but before we get too ahead of ourselves, you might've noticed the top says subscribe to start creating. And now we come to the unfortunate realization that creating images is expensive and we have to pay. So, click on join now to check out the options.
As of filming in late 2024, there are four plans, and they mainly differ in the number of images that you can create and how fast you can create them. If you're using fast generations, it'll take about 10 to 20 seconds per prompt to create your images. Relaxed generations, on the other hand, can take a long time because you're waiting in line with everyone else using relaxed generation.
The limited generations that you get on the basic plan are fast. A good rule of thumb is that one image generation is one minute of fast generation time. Bottom line, I choose between the basic plan and the standard plan for personal use or standard and above for commercial use.
Now, if you run out of fast hours, you can always buy more and add them onto your plan. We won't cover this today, but you can also earn fast hours by completing tasks like surveys or image ranking. So, choose your plan and let's go back to the explore tab.
Let's find a good image to play with like this Tim Burton character. As you can see, the prompts can get quite specific. Let's copy the prompt by clicking on use prompt.
Let's make a small change to make this a female character instead of a male character. Now just press enter to start the generation process. Now you can keep browsing here if you like, but let's go over to the create tab to see our results.
Congratulations, you've just created your first image with Midjourney, quite spooky, isn't it? I love that last one as a new AI avatar for the channel. Let's back up a bit.
Every time you submit a prompt, Midjourney creates four different images for you to choose from. And there is a setting we'll cover in a minute to vary them more. For now, though, let's create a new prompt and this is what I call a novice prompt.
We'll just type girl on a playground. It's not very specific, so we might not get something that we expect. Because the prompt wasn't very detailed, you give Midjourney a lot of creative license.
For example, is this a photo or is this a painting? What's the weather like? How many clouds are in the sky?
What is she wearing? Where is she looking? So, let's try a more specific prompt.
We're going to add what piece of equipment she's on and what expression she has on her face. Let's look through the results together. Click on any image to zoom in.
You can use the arrow keys to navigate between the images. Notice how even though we were more specific with our prompt, Midjourney still took creative license with a couple of things like what she's wearing, the feel of the image, her hairstyle, et cetera. Let's go back and talk about a few options.
We'll reuse the same prompt and this time click on the options icon in the prompt box. Lots of options here. Let's go through them one by one.
First up is image size. This affects the aspect ratio of the generated images. We can skew to portrait, landscape, or anywhere in between.
Let's go with a portrait view for our next one. Next up is the model. First up is the mode parameter.
So Midjourney AI has some opinions about what makes an image look good. So, if you leave this in standard mode, Midjourney will apply those opinions and optimize your image. Otherwise, if you switch this to raw mode, Midjourney will leave your image alone.
Raw is usually better when you have very highly specific prompts. But this could change with different versions of Midjourney, which leads us to the next option where you can explicitly control what version of Midjourney you're using. If you just started using Midjourney, I'd leave this as the default.
Personalize is available once you've ranked a certain number of images in tasks. Midjourney will personalize your images based on the aesthetic that you prefer. Now on to some of the fun ones in the aesthetics category.
First up is stylization. Remember I said that Midjourney has some opinions about what makes a good image? Well, you can scale that up or down depending on how artistic you want to make your result.
Let's see this in action before we move on. We'll crank the stylization to zero and start a new prompt. Let's make it a child's drawing of a dog.
We'll hit Enter here to start the generation process. Now let's crank the stylization all the way up to 1000 and then type the same prompt. Let's see what comes up.
Okay, let's get a look at these. Here are the very artistic Midjourney predefined images. One, two, three, four.
And here are the less artistic ones. One, two, three, and four. What you prefer depends on your mood.
Let's go back to the settings. Let's move the stylization back to the default of 100 and then talk about weirdness. The best way to explain weirdness is to show you.
I find it works best with prompts that are more open to interpretation. So, let's make a new prompt here. We'll go with a peacock with pink, white, and green shiny feathers.
Please excuse my lack of commas. We'll keep the weirdness at zero for this one, but crank the variety up to 100. The greater the variety, the greater the difference among the four images generated.
I'm doing this so we can see multiple weird options. Let's submit the prompt and then submit the same prompt with weirdness set at 1500. Nah, just kidding.
We're going all the way to 3000. Now this is fun because this is the reality of AI image generation. In the non-weird version, we didn't get any peacocks.
Well, maybe you ornithologists in the audience can correct me on this one, but this doesn't look like a peacock to me. Now here are the weird ones. I don't know what the heck's going on with this peacock's head, but I think all of them are pretty cool.
The last option we didn't cover is the speed. Now the Midjourney docs say that turbo mode generates images up to four times faster but consumes twice as much subscription time. Now let's say that you really like an image and you want to refine it in some way.
Let's take this girl on the slide, for example. In the bottom right, you can see a number of actions that you can take. Click on more options to see all of the available possibilities.
You can show or hide them depending on which ones you like. The vary option will make similar images to this one that are either minor and subtle or strong and sweeping. Upscale makes a higher resolution version of the image.
Remix allows you to modify this image with a prompt. Pan adds more content in the direction of the arrow that you click on. You can also zoom in, rerun the prompt, or edit specific parts of the image using a prompt or not.
And finally, you can use this image, the style of this image, or the prompt itself as the basis for a new prompt. Let's try a whole bunch of options and see what we get. I'll vary this subtle, strong, and I'll upscale this creatively, and I'll remix strong.
Here, I'll change this to be an orange slide instead of a yellow one, and press enter. I'll also pan this up. Let's also try the editor mode.
Here, we can erase parts of the image that we want Midjourney to regenerate. Let's erase everything except the girl to see what happens. We'll increase our eraser brush size to the maximum, and then click and drag around the girl to hide the slide.
That's good on the left side. Let's get the right side and the top. This won't be perfect, but let's see what happens.
Let's exit out of this and look at our creations. I'm all for the slow reveal, so let's check these out one by one. Okay, here are the very subtle images.
You see the hairstyle slightly changed; the backgrounds very slightly changed. Here are the strong variations. Notice the hand position has changed, notice the slides have changed, and in this one, we even see the bottom of the shoe.
Here's the upscaled version. You see hyper-realistic photo. Here's the remix where we asked for an orange slide.
Quite different than our original image. Here are the images we asked to pan up. Notice how some of these are not quite real.
And finally, here are the images where we erased the slide and asked for a region. We'll finish off this tutorial with some advanced prompting techniques, so let's go back to the Explore tab. You can use any image as part of your prompt by clicking on the image icon when you hover over another image.
And if you hover over the image in your prompt, you'll see there are three types of ways that image can be used. We'll start with the image prompt on the far right. You can combine multiple images or an image in a text prompt.
Let's combine this Tim Burton character with the Mandelbrot rainbow cereal. Press Enter to submit that one. Next, let's use an image as a style reference.
Click on the paperclip icon. Style references impact the aesthetic of the final piece. Let's add a peacock as our prompt and submit that one.
Finally, we can use the subject of another image as a character reference. Once the image is in your prompt, click on the person icon. Let's add a prompt, woman driving on a futuristic Grand Prix course.
The character from this image prompt should be driving in that car. Let's press Enter to submit. Let's look at our creations.
Once more, you can indulge my slow reveal here. Here's the Tim Burton character with the rainbow cereal kid with no cereal to be found. Here's that peacock and the aesthetic of the horse image.
I don't know about you, but this doesn't screen that aesthetic to me. And finally, here's that character on the Grand Prix course. Looks pretty good, but I guess because we kept our weird setting at 3000, she's now part of the race car.
But who knows, maybe that's how we'll be racing in the future. This is an awesome tool that gets me excited about image generation. So, share your favorite prompts in the comments to inspire our community to create some very cool art.
I'll see you next time.
