Guo shengwen

Midjourney is often open, but I don't know the graphics I want to generate. I don't know how to describe keywords.

Here, let's talk about the skills of a graphic essay recently published by China Travel Service:

Step 1: Enter/describe in the input box.

Step 2: Drag the reference diagram into the box;

Step 3: Enter, and M will generate four groups of keywords according to your reference picture. 1, 2, 3 and 4 at the bottom of the picture correspond to the numbers before the keywords.

Step 4: Click the numbers below to operate the map of Wen Sheng with the corresponding keywords, or you can modify the keywords yourself.

Step 5: Obviously, the pictures described by AI are still somewhat different and cannot be used directly. We need to screen out more similar pictures, such as 1 and 4, and then put them into the translator for further adjustment.

Make the following simple adjustments to make the shape and composition of the whole subject close. If we want to continue to approach the original picture, it is the continuous optimization and adjustment of style and color. The defects are obvious, too complicated and need to be adjusted repeatedly. It is generally used in conjunction with drawing operations (that is, drawing drawings).

U/V

What if we think the overall effect of the generated diagram is good, but we want to optimize the details again?

Take the following figure as an example, and fine-tune it on the premise of keeping the style stable. Select the U/V button at the bottom of the picture, and 1-4 corresponds to O-4 of the picture respectively. Here I choose to think that the first picture is good:

Clicking on U 1, /AI will enlarge your first picture. Click V 1, and AI will make partial modification according to the first picture.

Mixing function

The second point of fine-tuning can only be clicked with the mouse, AI can operate by itself, and remix mode can make some adjustments on keywords.

Mixing, as the name implies, is mixing. Add and mix some things on the basis of your original picture, and the overall shape will not change.

Step 1: Find the settings. You need to open Remixmode in the settings, which is Remix mode.

Step 2: Find the mixing mode.

Step 3: Select any deformation mode to open remix mode. Changing the mode of U/V will change the parameters, and the button will change from blue to green.

Step 4: Add the keyword blue.

Step 5: Compare before and after. The overall shape has hardly changed, and the blue decoration has been added, and the colors are more abundant.

Seed function

As we all know, the drawing operation in the middle is random every time, so how to output the same picture stably? The seed function is needed here.

Seed can be simply understood as the initial template of AI drawing, and the same picture can be generated with the same keywords and seed values. Similarly, you can use the same seed for keyword fine-tuning to achieve a relatively stable style. The following is the specific operation:

Step 1: Add a reaction

Step 2: Find the envelope, and there will be a prompt in the left column. If you don't search for envelopes in the search box.

Step 3: Copy the seed value.

Step 4: Use seed Ghibli-style spacecraft, pure interface, rendering style in 4d cinema, funny character design, orange and bronze Bloommore, toy core, solid background, clean background, movie lighting, super details, 8K, ultimate seed 2334569559.

Where-seed is the command to add a seed value (make sure to add a space between the seed and the value)

Image fusion function

Image fusion is the fusion of multiple images by AI to realize the blending of styles. (The upper limit of image fusion is five. It is better to upload images with the same width and height as expected, and the effect is better. ) The following is the specific operation of the function: Step 1: Input/Fusion.

Step 2: After dragging in the picture input, you will be given two picture upload boxes by default. If you want to do it, you need to click Add below, and then add multiple pictures will pop up above, add the option bar, and click the picture add picture upload box.

So how can this function be optimized in a stable style? My usual usage is to fuse two good but slightly flawed pictures of elements, such as the following one:

Referring to the picture, the overall effect of the first gift box is better, but the bow is a bit strange. The second style is poor, but the bow is very effective, so I will combine them.

Mixed result diagram

The effect of fusion is not necessarily complementary, but may be more unfamiliar, so it needs to be tried many times.

The difference between word combinations and sentences

Usually, when generating graphics, words and sentences are often mixed, and few people pay attention to their differences. After all, although Midjourney has weak Chinese recognition ability, it doesn't need so much grammar. How can I input AI to be recognized, so will the use of word combinations and sentences be different?

A little girl sitting in a spaceship in a spacesuit is sitting in the spaceship.

Topics are described as word combinations.

A little girl, astronaut costume, spaceship.

Obviously, if it is a sentence, the relationship between the little girl and the spaceship is very clear in the whole picture, and the little girl is sitting on the spaceship. But if it is a combination, the little girl may become parallel to the spacecraft. Therefore, the different arrangement and use of keywords may have different effects on the performance of the screen. You must try more to discover more wonderful uses of AI.