Transcription
Hello there. Grab a coffee and get cozy, as I break down everything you need to know about Midjourney in 2025. My name is Nolan Michaels. I've made over 77,000 images in Midjourney. I think you could argue about what you should learn first, so there will be time stamps below in case you want to skip around. Let's get started right at the beginning: how to sign up.
If you go to Midjourney.com, you'll be greeted with this screen. We can click down here at sign up. And for the longest time, we only had the option to continue with Discord. Now we can sign up with our Google account. This is going to be way easier, way cleaner. And for the purpose of this beginner guide, we're going to be doing everything through the Midjourney website. I don't think I'll be showing you anything on Discord in this video. I hope that's okay with you.
Now let's jump to the website and we'll talk about prompting. The Midjourney website will open up on the explore page. This is where you can see everybody's creations from the entire community, the public images that is. We'll get to how you can generate in private a little later on, but I want you to ignore all of this for now. Right at the top is the prompt box. That's where we're going to focus. This is where you can type anything into Midjourney and it will create a picture of that for you. And look, I don't think there is any wrong answer when it comes to the best prompting practices, but there are a few tricks you can use to help you generate the best results. And I'm going to be covering those tricks in this video.
The first philosophy I think you should follow is starting small and then elaborating after. Let's just go through an example so I can show you what I mean. Instead of writing a big long prompt all at once, I want you to start as simple as possible because we don't really know what Midjourney is going to create. We want to start small and then we can start adding in details after. So I'm going to start really simple. I just want to see a koala. And all I'm going to do is hit enter.
Over here on the left side, you'll see a little one pop up. This is the create page. If we click on that, it will take us to this page where we can see our images being generated. Every prompt you run, Midjourney is going to generate four options for you. Isn't that amazing? And here we go. We have our four images of a koala. We can click on one of the images to enlarge it. We'll get a better view of it. From here, we can scroll to see more, or we could use the arrow keys on our keyboard. If we click in the negative space or on this X here, it'll bring us back to the create page.
Now we're going to see my philosophy in action. We started small, we generated a koala, but these pictures are kind of boring, right? So let's start adding some detail. What did I really want to see from the koala? I want to see a koala dressed like a gentleman. What do you think it's going to create? Okay, now we're getting somewhere. Our koala is looking mighty fine in his new outfit. And do you see how I didn't have to describe the outfit he was wearing specifically? I just needed to provide a general aesthetic: "dressed like a gentleman," and Midjourney will handle the rest.
But now that I see these, I think I know what I want next. I want to change the setting. I want to change the background. This is still kind of boring to me. And instead of typing the prompt out again, I can actually scroll over to the right. We can rerun the prompt, but that's not what we want. We actually just want to click on "use." That will fill the prompt box with that exact prompt, and now we can make our changes from here. So I'm going to put a comma after "gentleman" and then I'm going to describe the background. I'm going to say something like, "the koala is standing outside of a bar."
Now here's another little prompting trick. What did I do here? I referenced the koala again. That's what is known as a callback. Instead of saying "the koala dressed like a gentleman, he or she is standing outside of a bar," I'm going to specifically refer to the koala again. And I'm doing this because the longer your prompt gets, the more Midjourney has a chance to get confused. We want to avoid all possible avenues of that confusion and we want to be as clear as possible. We're kind of writing it like instructions: "a koala dressed like a gentleman. The koala is standing outside of a bar." Try to use proper names when you're referring to different subjects in your prompt.
And there we go. We have four new options, and I'm liking these a lot. But now you'll notice something else: that not all four of them are going to match your vision. Like this one right here, sure, he's standing like away from a bar, but I kind of meant outside, right? And that's when you'll realize that Midjourney is going to give you four options because it's not always sure what exactly you're looking for. But like this one right here, I think that's a pretty awesome picture. But there's still more we can add to it. I'm thinking there's at least two or three more details of my vision that I haven't seen yet.
So again, we can go over here and click on "use." That will load up the prompt box and make it easier to make adjustments. I'm going to try saying something like, "the koala is smoking a cigarette on the street. The koala is standing outside of the bar." Again, I'm using callbacks to our opening statement: "the koala." All right, now we're getting somewhere. I'm not really a fan of one, two, or four, but number three here, like that, that is so funny. I like this one a lot. But there's still a little more we can add.
I'm going to add one detail to the street. I'm going to say, "smoking a cigarette on a busy street," or maybe "a crowded busy street." Maybe that will work. I just kind of want some more action in the scene. And then I'm going to add sort of an art style or a reference. I'm going to use one of my favorites: "a vintage Polaroid." And look at that. Look where we are now. How awesome is that picture? That one's really cool as well. Oh, I actually like that a lot. And that one's awesome. Ah, they're all great. Midjourney is the best. So much fun to use.
And I hope you're understanding why I suggest that you should start small. Instead of writing a long prompt like this all at once, start with the most basic version of your vision. So we'll go from "a koala" to one that's "dressed like a gentleman" to one that's "standing outside of a bar," then we have him "smoking a cigarette," and now we have "a vintage Polaroid of a koala dressed like a gentleman. The koala is smoking a cigarette on a crowded busy street. The koala is standing outside of a bar." Do not get overwhelmed by all the possibilities. You don't have to write long prompts like that right away. Start small and build your way up. This is a proven technique. It works so well with everyone I've tried it with, and it can really help hone in your vision because it will give you the time to add the details you're looking for. Trust me on that one.
I teach this kind of stuff in my Midjourney beginner course. It will be getting a full update for Midjourney version 7, which is expected to release near the end of January. My course is 25% off until that time, and then the price will be increasing as I film the new material. Check it out if you want a full, structured guide through every single feature.
I have another prompting trick that you can try that is super simple but super effective. It is essentially just deciding between what you mentioned first in your prompt: the subject or the setting. Your choice can make a massive difference. Historically, what you mentioned first in your prompt is what Midjourney is going to emphasize the most in the generation. Let me show you an example.
Let's try "a guitarist strumming under a dimly lit street at night." And before we had entered, let me show you how you can change the frame size of your generation. We can click on this little settings button over here, and you'll see that we're generating in a square ratio. We can change it to a portrait or a landscape, and we can take this slider right here and change it that way. Let's go with a 16x9 for now. You can hit the enter key, or you can click on this little arrow button they added up here, and you'll see that it's generating in this wide frame.
Now, we mentioned the guitarist first, the subject. And I think you're going to notice that in the results. Do you see how they're front and center, basically? Watch what happens when we make a slight change to the prompt. We're going to put the subject first. Instead of "a guitarist strumming under a dimly lit street at night," we're going with "a dimly lit street at night, a guitarist is strumming." And look how different some of the results are. Sure, number one and number four look kind of similar to the previous attempt, but I'd say two and three are quite different. That's because we put an emphasis on the setting first. First, when the subject is first, you're going to notice them a lot closer to the camera, usually front and center. Whereas when the setting is first, it's going to take priority, and then your subject can lay anywhere within that frame.
Let me show you one more example that will clearly illustrate this idea. Here we have "a person walking in a city park." You'll notice them kind of in the center of the screen. But check out when we put the city park first: "a city park with a person walking." Look how much farther back they are. Isn't that wild? Essentially, the exact same prompt, just switching around the subject and the setting. This might not seem like the biggest change, but this is how you can go about re-emphasizing the focus of your generation. Don't forget about this technique.
Now I want to talk to you about parameters. These are things you can add to your prompt to change the results. You can think of them like adding spices to a recipe. Some salt and a little bit of pepper will change the taste of a dish. And in Midjourney, there are a handful of powerful parameters. But the first one I think you should know about is called the stylize value. You can think of this parameter as determining the trade-off between beauty and accuracy. The value ranges between 0 and 1,000. At 1,000, we are telling Midjourney we prefer beauty over accuracy. And at zero, we are telling the bot we want it to completely focus on our prompt and ignore making it pretty. Those are two extreme ends: zero and 1,000. 100 is the default, which means that every time you write a prompt, Midjourney is going to use a stylize value of 100. You can increase it for more beauty or decrease it for more accuracy. But again, you have to remember that tradeoff.
Let me show you. Let's take a look at this prompt here: "a children's crayon drawing of an elephant and a penguin near a purple waterfall." This is at stylize 100. This is the type of look you would get if you ran this prompt yourself. If we lowered the stylize value to 50, this is the generation we would get. Very similar to the default, I'll give you that. But look what happens when we lower it all the way to zero. You notice how these are kind of ugly? That's because we asked it to follow our prompt exactly: "a children's crayon drawing." And like that, that is a child's crayon drawing. It's not very good, but that's what happens when we lower that value.
Conversely, look at what happens when we raise the stylize value. Here it is at 200, 300, and 400. You'll notice how similar they look. And that's why you should know that raising the stylize value doesn't matter too much. Right around 100 is where you want to be. You could raise it a bit or decrease it a bit without changing the results too much. But again, look what happens as we move towards the other extreme, the beauty side. Here's 500, 600, 700. But look at 800, 900, and 1,000. We're starting to tell Midjourney to focus way more on beauty than what we wrote in our prompt, which is why it's starting to stray from the prompt in the generation. Like the penguin is starting to turn yellow. There's no elephant in this one. And you'll notice that the pictures are way more beautiful than the lower values. That's really the tradeoff: more beauty and less accuracy versus uglier pictures but more accurate. It's a really powerful parameter, and keep that trade-off in mind.
Let me show you how you can use this parameter yourself. We can type our prompt and then we can trigger this value with a quick shortcut by typing `--s`. That stands for the stylize value, and `--` is one way you can trigger all parameters inside of your prompt. So we could finish this with the value between 0 and 1,000. Or we can click on our settings menu over here, and you'll see the stylization value right here with a slider that goes up by 50 at a time. So we could lower it to zero and then hit enter. Here's the dog and cat at stylize 1,000 and at zero. Pretty big difference, right? Quite beautiful versus not a very good picture at all.
The next big parameter that you're going to want to know about is called the chaos parameter. And this one is extra powerful because it determines how much variety you're going to get within your generations, as in, within the grid of images that Midjourney generates for your prompt. You can find it in your settings as well, under "variety." High values will produce more unusual and unexpected results, while lower values have more reliable, repeatable results. And that's an important distinction because the default value is zero. The Midjourney team knows that the chaos value is a little too powerful to be left on its own, but it is something you should be exploring.
So if we prompted for "a 1990s anime character, a futuristic knight in blue armor," this is using an aspect ratio of 2x3 with chaos zero, and this is what we might get. Sure, there is a bit of variety already within the grid, but they all do kind of look similar, at least these three do, with this one being an outlier. We can also type in our prompt `--c` that stands for chaos, and then a value between zero and 100. So watch what happens when we use chaos 3 for this prompt. We're starting to see that each of the pictures is starting to look a little different from each other. Not a lot of similarities going on. And that effect is going to increase as we increase the value. Like here, we are at chaos 6, and I honestly think that this might be even a little too high of a value. Like, I think you should stay around between like two and five, if I'm being honest. Of course, it's up to you, you can explore at your own will. Do not let any of my suggestions hold you back. I'm just telling you for best practices.
The higher the chaos value, the less likely you're going to find what you're looking for. But at the same time, you might discover a diamond in the rough. It really is up to you. Because here's chaos 9, and look how cool these are. Like, one of these might be what you wanted. I like this one a lot. So maybe chaos 9 is like the upper limit of what I'd suggest. Here it is at 15, and you can see that we're starting to lose a bit of the prompt. I don't think this one here is really blue, but I do like number one a lot. I think you'll start to understand what I mean when you see chaos 20. Like, maybe only one and three really resemble what we wanted, while two and four, not so much. Here you can see chaos 30, and we're starting to lose a bit of what we wanted. Chaos 40, and yeah, I don't really think you'd want this. And then take a look at chaos 50. This is the exact same prompt, remember that, and this is what we're getting. Like, these are pretty nonsensical, albeit cool, pictures. You're still going to get something cool, but if you want Midjourney to follow your prompt, high chaos values, it's not the way to go. And then to emphasize the point, here is chaos 100. Maybe it's got the anime look, but this is not a futuristic knight in blue armor. None of these are. Be careful with the chaos value. It's super powerful. I honestly don't recommend going above six, but of course, it's up to you. Feel free to explore. You really never know what you're going to find.
And I guess I have to apologize if some of my advice is sort of wishy-washy. Like, there are no set rules on how to go about using AI. It's really about exploring at your own pace and using the AI to amplify your imagination. That concept alone means there's nothing written in stone. Please don't let anyone ever tell you there's something you have to do. These are all just suggestions, but I do hope they help you out.
The next parameter that I think you should be trying is called style raw. It's a bit confusing, but it's essentially the exact same thing as the regular Midjourney version, just slightly more boring. It sort of takes your prompts more literally, and that can be really helpful in certain situations. You can style raw in your settings under the "model version." We have "mode: standard" and "raw." Alternatively, you could type `--style raw` at the end of your prompt and then the word "raw." This is going to invoke style raw for your generation.
And we'll take a look at this example: "a National Geographic photo of a jaguar." We're using a wide 16x9 aspect ratio. And when you run this with the regular Midjourney algorithm, this is what we get. Number three is a cool picture for sure, but then one, two, and four, maybe, yeah, these aren't bad, but I don't think they were really what I was hoping for. Style raw happens to excel at realistic images because it is more literal. So when you want something photorealistic, that's when I would recommend using style raw. Take a look at this. Aren't these way better? Like, way more of a National Geographic photo. This one, number three, is pretty similar to the regular version, but how about number two? Look how beautiful that is, all because of style raw.
And don't forget that you can use other parameters in addition to style raw. Like, if we wanted more beauty, we could raise the stylize value. And if we wanted it even more realistic, we could actually lower the stylize value. These kind of combinations can be really powerful. Don't forget about that. Let me show you another quick example. Take a look at this prompt: "a portrait of an ancient queen, opulent maximalism style, Unsplash." On the bottom, we have regular Midjourney, and on the top, we have style raw. Looking at the bottom, we can see that these are a little more artistic, like this one especially, it looks kind of like a painting. Whereas style raw is literally just a portrait of a queen. I know that number three is similar in both, but I think you can tell the differences in the others. Do you sort of see what I mean that the same prompt can have quite a different result depending on if you use style raw or not? Basically, if Midjourney isn't listening to you, please try style raw. It can help.
The last parameter that I think you should know about right now as a beginner is the seed parameter. And you may have noticed me using this in some previous examples. Basically, every prompt is run from a blueprint, and these blueprints are called seeds. So if me and you ran the same prompt with the same seed number, we would come out with the same pictures, most of the time, like 99% of the time. There are some little variations that happen, but don't worry about that for now. Just understand that seeds are your starting blueprint. When you don't specify a number, Midjourney is going to choose a random one. And there are over 4 billion seeds. That's why you're never going to see the same image generated twice. The odds are just extraordinary.
But again, we can use `--seed` in our prompt to pick the number for the bot. And why would we want to do this? Well, it's perfect for testing individual words. Let me show you. Let's say we were wondering about different facial expressions and which one we should use in our prompt. We can use seed numbers to test this perfectly. So we could write something like this. I mashed the keyboard for some random numbers. It really does not matter. Use your favorite lucky number if you want. But right here, I put a word inside of curly brackets and I started it with a comma. The curly brackets are how you use a feature called permutations. And permutations allow you to write more than one prompt at once. This can save a lot of time. And we write it like this: `{,word}`. That tells the bot to run this prompt with and without that word. When combined with a specific seed number, this is the most scientific test you can run in Midjourney. It's a super helpful shortcut. It's going to save you a lot of time. So it's going to generate "a woman" and "a happy woman." And then we can tell what effect "happy" has on the generation. As you can see, the results are quite different. This works really well for facial expressions or even colors.
So this prompt is going to run "a cowboy hat" because we have a comma at the start, as well as "a blue and a pink cowboy hat." And you see how all of these are super similar? That's because of that seed number. We locked in the powerful variable. We locked in the blueprint. And I guess I should just point out that sometimes you might come across these really long prompts, and you're probably not going to understand why certain words were used. This technique allows you to take certain words and test them individually to figure out what they're really adding to the results. This is a little more scientific, it's a little more advanced, but I do think you should know about it. I hope that makes sense.
Now I really want to show you the type of options you have after you find a picture that you love. And the one I really want to stress to you is called variations, specifically strong variation. Let's go back to the vintage Polaroid of a koala from before. And anytime you hover over a picture, you will see the "V subtle" and "V strong." When you click that button, Midjourney will generate four new versions of that specific picture. It will be like 96% the same, but still different. You'll notice how it looks really similar, but there are some big differences in terms of like the clothing. The details are different in each one, and maybe his pose is going to be a little different as well. If you were to click on any picture, you can also find variations here. Like I have "very subtle" and "very strong." If you don't see that, you can click on "more options" and it should be here. I would click on "vary."
And this technique would be known as "V rolling," hitting variations over and over again in hopes of finding the perfect image. Like, I'll hit "strong," but then I'll also hit "subtle" to show you why I don't think you should be using subtle. And for the record, you can click up here on a variation near the title, and you can see the picture it came from. This will allow you to go back to the original. Don't forget about that one. So here are more strong variations from that picture. And like, sure, those are all kind of cool, but let me go back to the original. And now let me show you subtle variations. Do you see why I don't think you should click on that? Because like, what is it really doing? I think those changes are a little too subtle for you to ever be happy with the results. But that is subject to change. Maybe the variations will work a little differently in the future. I just want to stress while I have your attention: use strong variation. I think that's your best bet.
So one of my philosophies is that you should treat your initial generation as a first draft. And any image that you think has potential, click "strong variation" on it. This is a practical solution when you have something like messed up hands. You're going to get that a lot with AI, and strong variation is one way to allow Midjourney to course-correct itself and alter what went wrong with the first picture. Like, we have "a portrait of a woman holding a vase in an Art Nouveau style," and you can see that her hand is kind of messed up, right? Well, by hitting these "strong variation" buttons, like I said earlier, we're giving Midjourney a chance to course-correct itself. It's going to change up the image quite a bit, but we're hoping it's going to fix one specific thing. And we might get results like these. And like this one, I think that is so beautiful. It matches the original style really well, and it fixed the hands. Strong variation can be one of your best friends. Try to remember that.
So we ran went through the variation process, and now we found a picture that we love. What can we do with it? I think you should be looking at the upscale options. This is going to create a larger file size, higher resolution version of that picture. Like they say here, it's ideal for sharing, saving, or printing. Again, if you don't see that here, you would find it under "creation actions," "more options," and then "upscale" would be here. Do you want subtle or creative? I'll show you both. But unfortunately, I don't know if there's a right answer. I mean, you'll see in a second. Subtle is going to try and keep the original image as close as possible, while creative might take some liberties with that image. The changes are going to be really small, but there still could be changes, maybe in terms of color or small details. So depending on how much you care about your original picture is going to determine which upscale you use. I'd say it's about 50/50 on which one works better. Here's subtle, and there's creative. That's the original. And then look how close subtle is to it, like identical. But maybe you didn't really care, and you prefer the look that creative gave it. Like, take a look at the smoke. Maybe you prefer this smoke more. I really don't know. It is completely a personal preference. You're going to have to play around with both.
I think the next feature I want to show you is probably my favorite thing that Midjourney has to offer, and it's called the remix feature. It is a little hard to explain, but remix is great when you love the style of something but want to change subjects. Let me show you a quick example. Let's actually go back to our anime character here, the futuristic knight in blue armor. You'll see that we have "remix subtle" and "strong." Again, if you don't see this option here, you can find it under "more options." We do not want "remix subtle," it's just not going to work very well. We want "remix strong." And you'll notice that this loads up the prompt box, allowing you to make changes. It says "remix" right here. And all you want to do, or all I suggest you do, is change one key thing in the prompt: that being like the color or the subject. In this case, I'm actually going to keep it as a futuristic knight. I'm just going to change the color of their armor. So instead of blue, I'm going to write "pink." Aren't these so cool? I mean, I like this one a lot. But you'll notice one important thing: it still has chaos 9 in the prompt. I know I said just change the subject, but you should also remove any chaos values when you're using remix. So let's try that again. We changed this to pink, and now we're going to remove the chaos parameter. And when I hit enter, we'll end up with these. Like, whoa, isn't that so cool? They're not all perfect. I don't think this one really works, number one, but two, three, and four look like they could fit in that same world as the blue knight. And that's exactly why this is one of my favorite features. I like to find a style and then use that style across a bunch of different subjects.
So I'm actually going to write a permutation, like I showed you earlier, that way we can run a bunch of prompts at once. We'll go `{green,purple,yellow,wooden}`. I wonder how that'll work out. Close it with a curly bracket. And when I hit enter, Midjourney is going to run four prompts for me. That's just one way to save time so you don't have to go through this and write each one individually. And then there you go. We have our green knight, our purple, yellow, and wooden. It turned out not that bad. And then when you see all of these characters lined up beside each other, I think you can tell exactly why this feature is one of my favorites. Hope you enjoy.
I know I mentioned earlier that you can use variations to clean up something like messed up hands in your images, but there is a more practical way to do it, and that is through the Midjourney editor. If we took our picture here, "a portrait of a woman in Art Nouveau style," we can see that her hand is just, it's not quite right. Over here on the right side is where we have all of our tools. Under "more options," and then "more options," we're going to see the "editor" button. Now, when you try this sometime in the future, I don't know if that button is going to be where it is now, but that's what we're looking for: the Midjourney editor. This will bring up a new page with lots of options, and it defaults to an eraser. So we can click on any part of an image, paint on any part of an image, and generate a new part of the picture. So all I want to do is select the area around her hand. And a pro tip is to select a bigger area than you think you need. Don't be too precise, it's not necessary. The bigger, the better. We can edit our prompt up here, but that's not necessary. We're just trying to fix her hand. When we hit submit, Midjourney should know what we were looking for. And then look what we have here. Not the greatest, not very good at all, but that one's kind of all right. If the first four don't succeed, don't be afraid to try again.
But we can also edit other parts of the image. Let's try and give her a different color hair. So I'm going to do my best to erase the parts around her hair, and then I am going to edit the prompt. And the way you want to edit your prompt is as follows: you want to rewrite the prompt as if it's describing the new scene that hasn't been generated yet. Instead of just writing "blonde hair," we're going to rewrite the scene. We're going to say "a woman with blonde hair." We're just going to insert that change into the prompt: "a portrait of a woman with blonde hair holding a vase, vase Art Nouveau style." And like that, that should work. I erased her hand as well. Hopefully, we get another roll at that. Oh, and then look how cool these are. That's amazing. I think that's pretty good too. That might be my favorite. And that one didn't work out very well, but whatever. We have the other options. Look at that. Like, isn't that awesome? Remember that we want to rewrite the prompt. I tell you, that's probably the most important thing when you're using the editor: rewrite the prompt as if you're writing the entire new scene.
Let's click back down here and go into the editor again. And what we just did was called inpainting. That's where we erase part of the image and we generate something new inside. But you can also do what is known as outpainting, where we can extend the frame and generate something over here instead, or above. There's quite a lot of options. We can even select the middle of the image and place it somewhere inside of the new frame. Inpainting inside the image, outpainting outside of the original picture. Like, we'll hit submit. We can even shrink the picture all the way down, like place it over here on the side, hit submit, and Midjourney is going to paint in all of this. Now, it's probably going to be weird. Like, I don't know why you'd want to do this, but it's available to you. Have fun.
Here are the first examples where we kind of had an even frame around the outside, and these look pretty cool. And then here's what happened when we placed the original down here on the right side. It remade another woman for the rest of the image. And I'll tell you, that's because we left this prompt as is. Like, if we went back into the editor and we changed this prompt a bit, and we just said like, "Art Nouveau style," we got rid of "portrait of a woman." Now when we try this, we might like what we see. Oh man, these are cool. I didn't know what it would do, but it sort of made these like collage kind of things. Like, there's our original, but it added more characters around in similar frames. Like, that's so awesome. Good job, Midjourney.
Now let's reset everything, and then I want to show you something up here on the top right. This is where you can select pre-made frames so that you don't have to guess on your own. Like, if we wanted to extend it wide, we could click 16x9, and that will extend it for us to that exact frame size. It's going to save you a lot of trouble. But here's another pro trick for you: when you do this, I suggest you erase the outside of the original picture. This is going to help it flow more smoothly into the new image. I guess I could just show you an example. So we hit submit, and then we erase the frame. Just going to do something simple like that. I went a little farther here than normal. I think you could get away with erasing just a little bit. So here's what it'll look like when we don't erase the side of the frame. I think it's pretty, I'm not going to lie. I do like those. But take a look at what happens when we do erase the outside. It flows way nicer into the rest of the image. Like, these are beautiful. How cool is that? Remember that trick. It's really important when you're trying to expand the frame with the editor.
Okay, we made it through like 70% of the main features. Now let's go over how to navigate the website. I do apologize if you think I should have taught this first, but nonetheless, let's go over everything. Starting with one key question: dark mode or light mode? When you first load up the Midjourney website, it might look like this, very bright and white around the edges. Down here on the left side, at the bottom, we can click on this little sun, and we can change this to dark mode if that's your preference. Personally, I like the dark mode, so I'm going to leave it like this.
And then here we are on the homepage, the explore page. We can see all of the different pages we have access to over here on the left side. If I actually zoom out on the website, it will give you a better idea. We can see the names right here. So we have explore, create, edit, personalize, organize, chat, and tasks. Let's stick with explore for now, and this is where we can see all public Midjourney images. By default, it's on random, so you never know what you're going to see when you load up the Midjourney homepage. I think that's kind of cool. Over here on the right side, we can randomize again and keep rolling to see new home pages.
And then when you hover over some of these images, you can see a little heart button right here. That's the like button. If we click up here on likes, we'll see all of the images we have ever clicked like on. It's like a personal saved image tab just for you, but these were all made by other people. You should know that if you were to click on the "hot" tab, you're going to see the images that have received the most likes. So these are basically the most popular Midjourney images at this moment. But we can also see the top images from today. If we click on the side at this little arrow, we can change from the day to the week, or even see which images have received the most likes this month. That's pretty cool, right? So this is your explore page. I personally like to leave it on random because you never know what you're going to find.
Because we can click on these images and see what words were used in the prompt: "black cats smoking cigarette, house in the background in the forest." We can also see what parameters they used. In this case, it was chaos 30, stylize 750, as well as this profile. We'll get to profiles in a little bit. So that's your explore page. We already know what the create page is. This is the chronological order of all the images you've created.
This edit page is something a little different than the Midjourney editor. It's basically an advanced Midjourney editor. In fact, this page will let you edit any image, whether it was made by Midjourney or not. Super powerful, but because it's so powerful, there are some restrictions. At the moment, right now, you only have access to this if you fall under one of three categories: if you've made over 10,000 Midjourney images, if you've had a subscription for 12 consecutive months, or if you have an annual subscription. Those restrictions are pretty tight at the moment, but we don't know how long they will last for. By the time you try this, maybe it will be open to everyone. So I'm not going to spend too much time on this, but I can still show you around.
We can edit an image from a URL or we can upload one from our computer. Let's try this picture of me, and then I'll show you one fun thing that you can do. Let's erase most of everything, and then up here in the prompt box, we're going to say "a man with lots of tattoos." Midjourney is going to give us four options, and I think you can already tell how powerful this is because these do look pretty real. Like, that is one of the silliest pictures I've ever seen, but there you go. That's the Midjourney editor, the advanced editor, I should say.
Okay, let's do something new here. We'll get rid of this because I also want to show you this retexture feature. This is something that should be available to everyone, in my opinion, because it is a lot of fun to use with your Midjourney images. In fact, let me show you just how to do that. So we'll go to our create page, and let's take this picture here. This is a Vogue photo shoot of a Gothic woman holding a hamster. Okay, we can click on the editor, but then we can see "open in full editor." That will bring us to this page again. We can do some inpainting, make some changes, that's fine. But let's go to the retexture feature. And then in the prompt, let's change up the style. Let's go with like a 1990s anime look. I added bold, saturated colors. Maybe that'll turn out nicely. So the idea with retexturing is that it's going to keep the shape of the original image, but just change the look of it. I hope that makes sense. Like, woah, you're starting to see this already. So we'll go from our original to these. That's pretty cool. I probably should have kept a hamster in the prompt so that it knew she was holding a hamster, because without that, it's just going to change it to whatever it wants to. But nonetheless, like these, oh my God, that is so pretty. Look how fun the retexture could be. Again, this is only available to certain people at the moment, but I hope it's available to everybody soon.
Okay, let's explore some other options. I'm going to save "personalize" for a little bit because there's a lot to go through there. So for now, let's skip down to organize. This is going to bring up your entire gallery, every picture you've ever made. And there's a few key things that you're going to want to know about. First, over here on the right side, I suggest looking through "view options." I think you might want to change the size of all these thumbnails because right now they're on small, so you can't even really tell what's going on in the picture itself. Medium might be okay, or you might want large. Of course, this will take a lot longer to scroll through all of your images, but it's still important to know about. And you see how these were made in a wide aspect ratio? We can tell that because we have the layout set to full. If we change it to square, every single image will show up as a square. I don't really recommend that because it can be a little confusing. So I like to stick with full, and then maybe medium-sized.
The next thing you might want to know about are filters. This is where you can isolate just certain types of images if you want. You can just show images that you have liked personally, which in my case will just bring up this blue guy because I haven't been liking a lot of images. Although I do recommend that. So let me show you how you can do that. We'll go to our create page, and then on here, you can click on a picture and click the heart in the top right. That's fine. But what's way easier is right-clicking on an image and then clicking "like." And you'll see that when we right-click, we have all of these other options as well, these quick options where you can upscale right from here, or remix, go to the editor.
Another thing you might want to do is click "hide," and that will hide an individual image. This is just a way to like clean up your create page, give you some negative space, and hide images you really don't want to see. But again, back on the organize page, we can even search by the hidden images, and this one would show up right here. Hiding images is how you delete on the Midjourney website, but you should know that nothing is ever really deleted. You might be a little bummed out about that, but that is how it works at the moment. It's just a philosophy from their team, they don't want anything permanently deleted, but you can hide any picture that you want.
We can also filter by just our upscales, and then we can see our koalas up here. You can filter by image size. This could come in handy for sure. Like, let's just see portraits, landscapes, or square images. You can even search by version. You want to see some version 3 pictures? A cyberpunk dog in version 3. This is, this is where Midjourney started. We've come a long way. You can also filter by published or unpublished. Now, this is where I'm going to talk about stealth mode. By default, on your Midjourney plan, everything you create will be public, which means they're all published. If that is not what you're looking for and you want some privacy, you have to pay extra for it. That was important for me. So by default, everything I create is private, it's created in stealth mode, and then if I want, I can go into an image and I can publish it, meaning that it's public. We can see that stealth image generation is only available on the Pro Plan, which is $48 bucks a month if you buy the annual plan, or $60 a month if you purchase monthly. Look, that's a lot.
of money, but that's the price of privacy. With Mid Journey, it's just one of their philosophies. They want an open community, but they understand that some people want privacy, so you have the opportunity to pay for it. I know that might rub you the wrong way, but I have to let you know that's just how it works right now.
And like you'll see over here on the right side, I'm randomly all the way down here on the scroll bar. Back in June of 2023, I think that's the last time I made some version 3 images, and that's because we filtered by version 3 recently. So you'll notice on this scroll bar, it actually goes in chronological order, all of the days you've ever made images on. You can find them here. So if we wanted to see what I made on May 27th, we can click on that and we can see these beautiful suits. Pretty cool, right?
Obviously, we want to go back to the top. And now I want to show you folders. These folders will give you a chance to organize your images a little better. Let's actually create a new folder, and then we can see here, what do we want to call this one? You know, let's go with "cool." Something really simple. All right, and let's hit "Create Folder." So our "cool" folder is here, and then we can select any image and we can simply drag it into the folder, and there it is.
Now, what I should really tell you about is how to select more than one image. You can click anywhere in the negative space and drag across multiple images. But the other thing you should know about is the Shift key. This will allow you to click multiple images one after the other. We can also hold Shift and drag across multiple images, and then now that all of these are selected, we can slide them over into the "cool" folder. Pretty easy, right?
And then let's create another one, and this one I'm going to call "Night." And now I want to show you what a smart folder can do. We're going to click this little check mark, and then we can enter keywords, and anytime we have used those words in our prompt, Mid Journey is going to fill this folder with those images. This is pretty amazing. So for the "Night," we're obviously just going to do "night," but we might also add "medieval" here. Just make sure any words you add are separated by commas. So when we hit "Create Folder," and then we click on "Night," we'll see here that it's a smart folder, and it's filled with every time we've ever used "night" in our prompt. This is going to go back a while, all the way to August of 2023. Isn't that awesome? Smart folders are super powerful. Don't forget about that one.
Before we move on, I also want to show you that you can search through your entire gallery right here. So maybe we just want to see "knights." Like, that's pretty cool. And again, then you can scroll through, and every time you've created a "knight," you can create a smart folder. But searching through images is probably going to be a lot easier.
Speaking of search, though, let me go back to the explore page to show you something. On the organized page, it goes through all of your images. But on the explore page, it will go through the entire community. So if we type "knight" up here, we'll see every image created by the community that has the word "knight" in it. Isn't that awesome?
But then, let's say we find one specific picture that we really like, like maybe this one here. It's a nice painted picture of that knight. We can actually click on this little search icon up here to go one layer deeper. If we click on this, we'll now see every image that resembles that particular picture, which is going to be like a painterly knight in a field of flowers. Isn't that wild? I know that's a little niche, a little advanced, but pretty powerful, and maybe you'd want to know about it.
And then when you're on your create page, the search feature will again take you through all of your images. I know I'm teasing you a little, but we're going to save profiles again for a little bit because that's part of personalization. So let's go back over here on the left side and let's click on "Chat." These are public creation spaces where you can create with other members of the community. Like here's a daily theme room. You can generate images and see what people are creating live, as well as talk to each other.
On the right side here in the chat, you can create a private room that has like a special invitation code, or maybe it's like a password that you have to enter. Yes, you have to use a password for a private room, but that way you can create with your friends. You can even have voice chat. I guess, like, that's pretty cool. Personally, I don't spend too much time here, but I know some people do.
And now I want to show you what the "Tasks" page has for you. You're going to see quite a few different options, and this is a place where they offer rewards in terms of fast hours. They'll do this a couple of times a year. They'll offer these really long surveys to gather data on the community, and they'll provide you with fast hours as a reward.
But then there's also this community feature where you can help rate ideas and help them prioritize. So you'll see that they give you five different options, and then you have to rank the importance of these features. Like tutorials, you'll give that a few points, and you have 25 points to assign across each idea. Create an educational discount program, like that's pretty cool. Highest possible video generation quality, improving coherence, better visualization for variation jobs. And what you're doing with these points is determining how important each idea is in relation to the other ideas presented. If you do this enough times, they'll get a good idea of what the community wants. And we can see that on the leaderboard right now, a detailed character creator is the number one requested feature. You can also submit ideas over here on the left side. Pretty cool.
And another thing that is kind of cool that you should know about is you can help curate the Community Front Page, the explore page. This is going to give you a bunch of options. I guess there's 12 images here, and you have to select which one you think should show up on the front page. This can be kind of tough to choose from, right? Like maybe I'd go with this one. And if you happen to really like an image, you can hold Control or Command on your keyboard and click on it. That will open it in a new tab, and you can see what prompt was used to create the picture. So if you do that enough, and you're in the top 2,000 people who do that over a single day, you will earn one free fast hour. That's 60 GPU minutes, which is about 60 fast generations. It's a good win-win reward system.
All right, I know the first thing we covered in this video was prompting and the different ways you can use words to generate specific results. But with Mid Journey, you can also prompt with pictures, and that's what we're going to be talking about next. Let's go to the explore page and then watch what we can do. We can click on any picture and drag it into the prompt box. By default, this will add it as an image prompt. You'll see here that we have some other options, but we'll get to those in a bit. Right now, I want to show you how image prompting works. It's a little tricky to understand. It's best just to show you some examples.
So if we prompted for a car, Mid Journey is going to take this image and this word to create the final result. My guess is that we'll get a translucent car on a pink background. What do you think? Like, yeah, something like that. Image prompting can be a little weird for sure. As you can tell by this, it's basically like a dish or like two parts of a sandwich, peanut butter and jelly. Your picture plus your prompt is going to equal the results. And because it works like that, it's not very scientific. I don't really recommend image prompting too often, but it is cool. It can generate some really interesting results. A picture like this, all you needed to do was add this image plus the words "a bird."
But while we're on the subject of image prompting, there is one parameter you might want to try, and that is image weight. So let's add this image back to the prompts. We'll type "a car," and then check this out. I'm going to add `--iw`. That stands for image weight, and then we can use a value between 0 and 3. Zero would mean there's no influence, so we don't want that. And one is the default. I'll show you what two and three might look like. I'll be honest, I don't think you want an image weight of three because this will put the maximum influence on the picture rather than the words, and we're basically just going to get the exact same picture we started with. You're not going to see a car. Even an image weight of two might be a little too strong. So, you know what? Why don't we change it to something like 1.5? And there you go. These are a little nicer. Definitely not two or three, but one's not bad, and 1.4 is kind of cool as well. Maybe you'd even want to lower the image weight just a bit more, like 1.3, 1.4. Feel free to experiment.
So when you're on the explore page and you want to use one of these images, again, you can click and drag. But you can also use this button right here. This will add that image to the prompt. Let's say we find another picture we want to add, like maybe this woman right here. We could click on it, and then down here at the bottom, we can see "Use." We can use the image, the style, or the prompt. We'll get to the style in a bit, but right now, we could just click on "Image," and that will add it to the prompt as well.
Well, for image prompting, you actually don't need to include any words, which is kind of cool. But the other thing you should know is that you can add as many images as you want. I don't know if there's a limit. So we'll just drag a bunch. And I don't know if I mentioned this earlier in the video, but this is like one of the most important workflows ever. If you hit the up arrow on your keyboard, it will reload the last prompt for you again. If I hadn't mentioned that already, I apologize. That's really important to know. It's going to save you a ton of time.
So here's our woman plus the umbrella, and that's a pretty standard generation, right? And then here's the mix of the five different pictures. You really never know what you're going to find. You can also add different parameters to your image prompts, like `--stylize`, `--chaos`, `--style raw`. Everything goes. There is a lot to explore.
The next thing I want to show you is how you can put an image in your prompt and use it as a style reference. Like take this image of the umbrella. If you look closely, it's actually sort of like hand-drawn in a way, maybe like pencil crayon. So if we put this picture in the prompt and then click this middle button right here, that's going to add it as a style reference. And then in the prompt, we can write anything we want. Like, let's try "a car, a woman, and a sunset." Mid Journey is going to use this picture in the prompt, but it's going to be more like multiplication rather than addition. You can expect to see the same colors and style of creation, but not necessarily an umbrella. I hope that makes sense. Like, take a look at these. Aren't those so cool? Oh, I just got goosebumps. So beautiful. And if we look closely, we can see that this image is sort of hand-drawn in the same way as the umbrella. See what I mean? And it's the same for every subject. Style reference is super powerful. This is a great way to create consistency across multiple generations. You're probably going to use this a lot.
And the parameter that goes with style reference is called `--sw`, and then you can input a value between 0 and 1,000. 100 is the default. So if you wanted a little less influence of the picture, you could lower the style weight, maybe around 50. You don't need to go all the way to zero. And if you wanted more influence, maximum influence, you could raise the value all the way up to 1,000. Here's the car at style weight of 50, and as you can see, it's still blue, it's still yellow, it's following the same color palette, and it still has basically that same hand-drawn look. Here's 500, and here's 1,000, the maximum influence. And I think you can tell that the image might be having a little too much influence because these pictures are not that good-looking. We might be putting too much emphasis in one place. The default is the default for a reason. I'd recommend leaving it around 100, or even lowering it. Like, like that's a beautiful picture. Style weight of 50. Again, you're going to have to mix and match, play around with the different values. You never know what you're going to find. And I have to remind you, you can use every parameter with style reference: `--chaos`, `--style raw`.
AI is capable of a lot of things. There is a lot to play around with. Like, we'll put this here. We'll go to the explore page. Let's take this really modern-day, clean lighting picture and we'll put it in the prompt box. We're going to choose "style reference" for both of them. And then we can put, let's go with "a lightning storm." Right now, there is no way to determine the strength of each picture. It's just going to be 50/50. As a longtime Mid Journey user, that is something we're definitely hoping for because you can adjust those values on Discord. Since we're just focusing on the website right now, there is no way to change the influence. And look at these. That's pretty cool. We're getting like a realistic picture with that blue tint. Oh, that's awesome. And then maybe this is on the other side where we're getting more of a stylized drawing.
While we're here, I want to show you something. If you hold Shift on your keyboard, you can actually select more than one of these options. So we can have both images as both an image prompt and a style reference. We're going to get to this third option in a second. And what do you think? Again, there are a lot of ways to create really unique images.
All right, let's get to maybe the most fun feature of all, and that is character reference. Full disclosure, it does not work well with real pictures, as in your face. It's just a limitation of the system. It works better with Mid Journey images, but they're hoping to improve that in the future. So as of now, January 2025, you're only going to want to use this with Mid Journey pictures.
If you click this button next to your prompt box, you can see all of the images you've ever uploaded to Mid Journey. You could drag a file on here or choose one from your computer. I'm going to add this picture to the prompt, and then I'm going to select "Character Reference." This button right here. And then we can type a prompt, and this character will show up. So let's try something like "a woman in a chic fashionista outfit." But watch what happens. It didn't work out very well. She's still wearing that gray sweater from the original picture. I mean, it is her, it is the character, but that's not what we wanted. So we have to use another parameter. Let's use this prompt again, and then we're going to use the character weight parameter `--cw`. This has a value range between 0 and 100, with 100 being the default. And when it's at 100, it's going to take everything it knows about the character, including the clothing and hair. But because the system is not the greatest at the moment, it doesn't do a very good job. So I suggest lowering that character weight all the way to zero. I mean, maybe you can get away with 10, 15, 20, but I personally like going to zero. And when it's at zero, it's only going to focus on their face. And now watch what happens. We're going to see our character in the different outfits. How cool is that? Character weight, remember it. It's super powerful. Like, isn't that awesome? I mean, her hands are kind of weird, but I think you know what I mean.
Character reference also doesn't work very well with like cartoon characters at the moment, but it does a pretty good job with humans. Again, Mid Journey made humans. It's not going to work if you upload your own face.
And now that I showed you the three references: the image prompt, the style reference, and the character reference, I want to show you the triple triple. It's just like a little fun technique I like to use sometimes to create the most unique pictures possible. Let's go back to the explore page because, why not? We'll throw this person up there. Let's use the cat, and yeah, let's use this cartoon. I think that could be cool. We're going to hold Shift and select all of the references. This is a triple triple, three pictures with three references each. And then you have to write something silly in the prompt. It doesn't have to be anything, nothing specific. I like to go with just, yeah, it needs something. And then there is no way to know what Mid Journey is going to come up with. I promise you that. Okay, these are actually kind of gorgeous. I think I like that one. Oh, maybe this one too. That cat's pretty cool.
It is now time to get to Mid Journey's most powerful feature: personalization. This is your chance to modify the behavior of the bot to give you exactly what you're looking for. We can find it over here on the left side under "Personalize," and you'll be presented with two options. You can create a standard profile, or you can create a mood board. Let's go over the mood board first because I think that's what you'll want to know about the most. We can make a collection of images to show Mid Journey exactly what we're looking for. And I've made all of these. They are available to purchase, but you can get them at no extra charge if you join the Future Tech Academy, my community on the school platform.
So I've created this specific mood board called "Unsplash," and it's got all of these looks inside. It's about 20 images, and Mid Journey is going to take the average of all of those pictures when it's generating something new. So here we can go to "Copy Code," and then we can write our prompt. Don't worry about this for now, it's going to shorten to a smaller code after. Let's just go with "a turtle," and we can see how this turtle was generated in that style from the mood board. We didn't have to say anything else. You didn't have to describe the look. You just needed a predetermined board with similar styles. Isn't that awesome?
And if we look over here, we have the mood board "Unsplash." If we were to click on this, it will bring up this specific code. So that is my gift to you. If you want this exact look, you can use that code in your prompt, and it's triggered by `--profile` first. Now, to do this yourself, you can click this button right here, "Create Mood Board." Then we have the chance to add images from our computer, links from around the internet, or from our Mid Journey gallery. If you click the gallery, it will take you to your organization page where you can then search through your images. At the top here, actually, remember this style reference from before, the blue and yellow umbrella? Why don't we use that? Let's make a mood board out of this. So we'll just select a bunch of the images. We'll call this "Blue and Yellow." And now, instead of needing that style reference code, we have a mood board for it. Literally, all we need to do is hit "Copy Code," and then we can write stuff in the prompt. Let's go with "a car, an alien, a woman." Isn't that wild? Isn't that amazing? Sure, style references are really cool in their own right, but so are mood boards. Like, this might be a way easier workflow for you. It saves nicely on that personalization page. It's really easy to find again, and it's really easy to share. We'll just click on here, and here's the profile code. If you want to get back to that mood board, you can notice this little arrow right here, and that will take us to the board. And we made that with only seven images. Like, do you see how powerful that is?
And then what's really cool is if we go to the personalized page, we can see our "Blue and Yellow" board right here, and we can actually just click on "View Creations," and that will show us everything we've made with that specific mood board. And then another really cool feature is that if we were to make more images, like let's say "a dog," they'll show up right here. So you can basically create within the mood board itself. I think that's pretty great.
Let's take a look at another board. Let me show you this pop art cartoon look. I think this is pretty intense. And now I want to tell you about the parameter that will help with your personalization. This might get a little confusing, but it is the stylize value. We learned about that earlier, `--s`. When personalization is on, when you're using a profile, stylize controls the influence. 100 is the default. It goes all the way to 1,000. I'll show you what that looks like. So here we have the default, an alien on that mood board, and like, these pictures are so cool. I'm just going to favorite this one now so I don't forget about it. And then here's what happens when we lower the stylize value a little bit. I think you can notice it's just slightly different. Meanwhile, here's what happens when we raise the stylize value. The mood board becomes more important, and everything gets a little more bold. So if you create a mood board but you're not really feeling the results, consider raising the stylize value to give more influence to the mood board. That's just one little trick you're going to want to know about. And every other parameter also works: `--chaos`, `--style raw`. I've said it before, I'll say it again, there are lots of options to go through.
Back on the personalization page, we can also create a standard profile. This is sort of like a mini-game that you get to play and you get rewards for it. So what it's asking you to do is to select between the left or the right, which image do you like more? Now, I want to stress, don't worry if you don't like either of them. Just choose one. Choose the picture you like more. If you do this enough times, Mid Journey is going to learn what you like. It's pretty sophisticated, and you're able to create custom profiles this way. You actually only need 40 to unlock your first code, even though more rankings will help Mid Journey hone in on your style. Like, we choose this one, and then I like this one, we'll go that. So you do that 40 times, and you'll get your own code.
For my global profile, I've actually ranked 10,19 pairs of images, 10,020. So if I were to use my global profile, it's going to show up like this, but again, it will shorten in a minute. And then we say "a cat and a dog on a cute date." My global profile at the moment is this code right here, and this is the type of style it creates. I think this is very strange, number one, because it doesn't look like the other ones. Even though I like this a lot, based on what I know about my profile, I would expect this cartoony look to show up. So that is kind of strange, and maybe that's why I don't recommend the standard profile creation. I think mood boards are the much better option.
Lastly, I want to introduce you to Patchwork. This is Mid Journey's experimental storytelling feature. It's a bit of a world-building tool, and I want to stress that it is experimental. You can create directly on this infinite canvas, then you can generate plot points, characters, settings, items that they'll use in their world. It integrates large language models to help you write your story. It can be a lot of fun. This is a really cool feature, but there's a lot to go over, and I've actually already created a deep dive. So if you're interested in learning more, you should watch this video.
Now, I hope you're doing well. Take care, and I'll see you next time. Peace.