When you were still doubting and hesitating about AI, I had already spent tens of thousands of yuan to buy mainstream AI painting/video courses and paid communities on the market, and invested a lot of time in learning and practice. Through continuous summarization and integration, I have extracted a set of the simplest, most direct, and most effective learning methods.

If you want to master AI videos, mastering this set is enough!

01 Midjourney

In the first two chapters, we learned about the two major domestic AI painting software giants. If you haven't learned yet, you can check out the previous articles.

Today we are going to learn a painting software that is currently recognized as the number one.

midjourney

This is an all-English software, so are you ready? Let's get started

Let's set the tone for today's explanation. I will explain it from the perspective of practical painting, gradually introducing it, so that everyone can better understand. Otherwise, if I explain the concepts module by module, you will probably fall asleep before I finish

02 Basic Process

Midjourney: Login website: https://www.midjourney.com/

First of all, at the top, this text box is where you enter the prompt word. Here we first enter the prompt word and try to draw a picture

I entered the simplest prompt word, then pressed enter, and it was OK. At this point, we saw on the left side that create showed a picture at work. Create can be understood as your console. Our drawing homework is in this option bar, where you can see the work you drew

03 Immersive Learning Parameters

We click create and see this picture. You see, the simplest painting process is completed, and all subsequent operations are based on this. Let's take a step by step. If you don't want the horizontal screen size and want to draw a vertical screen, how do you operate it?

Click the icon on the right side of the prompt word text box to enter the following page

You see, the first column is the image size. We have three ways to choose.

First, we can directly choose the corresponding vertical, square, horizontal, and each template corresponds to a size.

Secondly, if the template does not have the size you want, you can also choose to manually drag and drop the Progress Bar below to adjust.

Third, you can also click on the value on the left to make changes directly, isn't it convenient?

I want a more realistic style, how can I tell it? No problem, it understands you

This mode has two options. If you want the picture style to be more realistic, choose raw. If you want the picture to be more artistic, choose stangard. The specific version depends on your work. The following version is relatively simple. For daily drawing, choose the highest version 6.1. If you are drawing anime, choose the niji6 version. Below is the personalized selection, which is commonly known as --p by the boss. I did not unlock it because I use a shared account. Don't mislead everyone's creation. By the way, what is a shared account? Take a look at this picture first

You see, to use this software, you not only need to have internet magic, but also need to purchase a membership. We need to buy at least $30 a month

04 Three major parameters

Let's focus on these three parameters this time

Stylization value:

The value range is 0-1000. The lower the value, the higher the term weight will be. In other words, the image will be generated according to the prompt words you wrote as much as possible. The higher the value, the lower the term weight will be, which will generate more artistic works.

This suggestion is to try from low to high first, so that the control is better

Weirdness strange value:

The value range is 0-3000. The lower the value, the more normal the work. The higher the value, the more free the work will be, giving you an unexpected impression

This suggestion is not needed for now. It is difficult to control and is equivalent to an advanced function. When you are proficient in drawing and have strong control over the picture, then try to use this function

Variety of:

This is relatively easy to understand. MJ generates four images each time, and this indicator refers to the similarity of these four images. Simply put, the lower the value, the more similar the styles of the four images you generate, and the higher the value, the greater the difference in style between the four images you generate

For this suggestion, in the early stage, you can adjust the style a bit to see which style you like. When you make minor adjustments, adjust the value a bit lower. Finally, there is an option called Speed, which is the simplest and refers to the speed of drawing. Generally, we choose fast. When the membership time runs out, it will automatically be reduced to relax. It is not recommended to choose turbo. Although the drawing is fast, it consumes twice the membership time each time, except for wealthy friends. We have set the basic parameters and will try to draw another picture

I saw that these four pictures have quite different artistic styles. If I have a good impression of the second picture of the kitten and want to make a few more similar ones, how should I operate?

You see, when you move the mouse over the corresponding image, a Vary Subtle and Vary Strong will appear. The former is fine-tuning, and the latter has a larger adjustment range. Here, it is recommended to try strong first to see more possibilities. When you find a more suitable image, you can use subtle fine-tuning. When we click in, we find that there are more options. What does this mean? Don't worry, let's take it slow

When you click in, you will find that the options for fine-tuning are no longer on the screen, but have been moved to the right side. The first row is fine-tuning. When you have fine-tuned and obtained the work you are satisfied with, what else do we need to do? Zoom in on the high definition, which is upscale zooming, still giving you two choices, Subtle, which almost maintains the original image. Creative will make some major changes on the original basis. It is recommended to use Creative for this part. Don't endlessly try creativity. If you are not satisfied with the work, you can choose rerun and let it rerun another set of images. If you think the picture is already good, but I want the kitten to wear a hat and regenerate the image of the kitten, then you need to use the editor function, the image editor, this function Super awesome, demonstrate it.

We click editor to enter this page, use the smear function to smear on the kitten's head, then write the prompt word on the right, and finally click generate in the lower right corner

We saw that the kitten successfully put on a Christmas hat, but it seemed strange and not very harmonious. It felt like the space above was a bit smaller, no problem, there is also a function of expanding the picture here

We dragged this small icon and dragged out some space above the kitten. Isn't the effect great? Look closely at the previous picture. There is also an option to expand the picture proportionally on the right. Imagine that you have drawn a very beautiful vertical work, but when applying it, you find that you need to use a horizontal version. It is difficult to maintain the same effect when reproducing. At this time, you can directly use the expansion function. The main body remains unchanged, but the expansion area is automatically filled. What else can you say? Perfect

If you are satisfied with this picture and want to draw similar works, what should you do? Let's look at the last option, which gives you more power than you want

You can use images, styles, and hints.

Using images, similar to mats,

Use style, similar to sref,

Using prompt words, you can refer to, modify, or even re-create.

This part focuses on explaining, because it involves a very important function, style consistency and character consistency, and also talks about the pad image and the difference between the two Let's first click on the image in use

After clicking, it will automatically run to the upper left corner. There are three small icons in the upper left corner. The little person represents character consistency, the paper clip represents style consistency, and the picture represents the pad image. You can hold down the shift function key and select two or three at the same time. What's the difference? Explain it Character consistency: Use Character consistency will make the subject of the new picture as similar as possible to the subject of the reference picture. You can also adjust different values from 0-100. The larger the value, the larger the reference range, such as body, clothes, etc. The smaller the value, the more it only refers to the face. How to use it? Add --cw 50 after the prompt word. Note that you must first type a space, and then in English format The --, then a space, and then the value of the 0-100 range, the format must not be wrong, otherwise you will find that there is no effect, we demonstrate that we refer to the main body, and then rewrite a prompt word

The result is like this, the size of the hat has changed somewhat, which we can debug by adjusting the cw value

This time we added a cw value to try it.

The hat has become smaller this time. This function is still very useful for maintaining character consistency when making videos.

Style Consistency: Next, let's talk about Style Consistency Using the style consistency function, the new image will be generated according to the style of the reference image. It can be controlled by numerical values, from 0-1000. The smaller the value, the better the control. The larger the value, the more free the style. Let's demonstrate that we choose a different style, which looks more obvious. We click on style, then look at the icon of the paper clip, and then write the prompt word. Here, I set the sw value to 100 to see the effect

It can be seen that although our subject has changed, the style can be seen at a glance, it is still the same style, and the specific can be adjusted slowly by adjusting the sw value

Pad diagram

Let's talk about the meaning of pad map pad map, which is equivalent to giving a map without telling it what to refer to specifically. Let the new map refer to this map intelligently, so the newly generated map will be very similar to the original map. There is also a value that can be adjusted, -- iw, which is the reference weight, with a value range of 0-2. The smaller the value, the smaller the reference, mainly relying on the prompt word. The larger the value, the greater the reference value, mainly relying on the reference map generation. The default value is 1. Let's demonstrate it

You see, the style is very similar. I didn't write a Christmas hat, but it still refers to a Christmas hat

05 Three points

I spent a lot of space talking about this part because it is very important. When we make AI videos, we will frequently use these functions

So to summarize:

First, try not to mix the three as much as possible, because then your control over the picture will be very poor.

Second, the uncertainty of the cushion map is relatively large, and the actual application is not as large as the first two

Thirdly, when you refer to a method that cannot produce results, you can try adding another one.

You must learn this function, otherwise you will find it difficult to make videos

This reference function can not only refer to your own work, but also refer to other people's excellent works. Demonstrate it.

We click on explore on the left, place the mouse over the image you like, and you can see four small icons. The first one is for reference. Click on the image icon first, then go back to modify the reference content. The second icon is a prompt word. Click once, and all its prompt words will be automatically entered into the text box. At this time, you can follow the prompt word to create the image or modify it for re-creation. The third search icon is to search for similar works, and the fourth heart is to mark the works you like

06 Organizing works

There is another function to briefly introduce. When you finish drawing too many pictures, how to classify them by category? We click organize.

He can create folders for management, and he can also quickly find why it is a simple introduction based on conditions.

Click on the picture, find the icon for direct download in the upper right corner, be sure to click on it, otherwise the format you downloaded is wrong.

07 Summary

Well, this is basically the key function of the midjourney web version. Why didn't I talk about registration? Because many friends use shared accounts. You don't have to worry about registration by yourself. Just log in to the account directly. This web version is newly released. New friends can use the web version directly. It is simple, convenient and easy to use. We used to use the discord version, and we also needed to memorize English instructions. It was very annoying. If an old friend is still using the discord version, let's take a look at another discord tutorial. Today's one is probably enough for everyone to digest it well. As the saying goes, knowing does not mean knowing. In practice, knowing how to use it is really knowing how to use it.

Another way to create graph is using AI tools from web, such as
AI graph maker
Graph-maker.ai