Discord is a very popular new chat tool that was born a few years ago.

Midjournal is used by sending the corresponding text to the Chatbot in the Discard channel, and the Chatbot returns the corresponding picture.

Join Midjournal

Register Discord account:

To use Midjournal, you need to first register a Discord account, open the Midjournal website at www.midjourney.com, and click "sign in". Use your Google email to follow the steps to fill in the information (if you have already registered an account, clicking "join the beta" will automatically enter "discord, new users do not need to handle this)

Then enter the Discord channel of Midjournal. After registering our account, we can use Discord in our browser or download their client. You need to connect the Midjournal server to Discord and enter the "Discord" interface. The specific steps are as follows.

Create Server

Click on the plus sign on the right and a pop-up window will appear for creating a server. Then, click on "Create it yourself" and select "For Me and My Friends Only". The server name and avatar can be the official default (optional), and the following information can be filled in freely. After filling it out, enter our server, and the creation of the server will be completed.

Add Midjournal Robot

After creating our own server, we cannot start drawing because our Midjournal robot has not been added yet. You can understand this as a group chat server that can bring friends in to chat and digress, Returning to our topic: Click on the "Explore Public Servers" button on the left side (the functions in the left row are all server functions, including the server where "I" is located, the creation server, and the public server. These functions are easy to understand and interested in researching and researching on their own). You should see our protagonist here, right? It's the first Midjournal, and then click on it (the welcome pop-up appears, let's ignore it).

A lot of information has appeared, and this community is public. Many people are working here to create images. The left column is the server channel, which includes the Midjournal management channel, posting some updates, and the public's image channel. I suggest you check out how people create images for keywords in it, but it is not recommended to generate your own images here. There is too much information.

The next step is to join the Midjournal community, which will also allow you to verify. We will click on it step by step. Note: After verifying it again, you may not have joined the community. This may be due to network reasons and the impact is not significant. Just repeat the verification process (ignoring those that have been added once). Below is the interface after successfully joining:

Then we switch to the channel, click on any one on the channel side, and then click on the "Member List" button at the top right (note that you need to click on one of the channels below to display the "Member List" button):

Click on 'Midjournal Bot', click on 'Add to Server', select the server we just created, and then authorization verification will be done. You can enter our server and see that the robot is already on our member list. This completes the task.

How to use Midjournal to create your first AI work

Keyword Generation Image

Return to our server and enter "/" in the input box below. A series of commands will pop up (if there is no pop-up window, log out and log in again). First, click "/image".

Enter a group of keywords. Remember: after a group of keywords, you need to add the English comma "," and then press the Space bar. Just leave one space blank. Look at my following group of keywords:

An mascot robot, smiling, modern robot, round robot, carton, flying, first up, crypto Coins background

Then we move on to the previous step and enter this set of keywords in the dialog box. I suggest that everyone not directly copy them, but instead use the keyboard to experience this operation.

Click 'Enter' to send, and then it will ask us for authorization. We will click 'Accept to' to authorize.

Then wait for the robot to produce the image, and the image will come out later:

Image instructions and exporting images

After the generation is completed, two rows of buttons will appear. U means to enlarge the image, U1/U2/U3/U4 means to enlarge one of the four images, and V means to use the composition of the image to regenerate a set of similar images. The order of V1/V2/V3/V4 is the same as that of U

For example, if I want to export the first image, we can click U3. Then, we can click on the image, open it in the browser, and save the image. The image will be downloaded (be sure to open it in the browser before saving it, otherwise the size of the image will be very small and blurry)

If you think the image is okay and the composition style is good, but there are some details that need further optimization, then we can click V1 and the next steps will be the same as above, isn't it very simple.

In general, we can generate multiple sets of images for the same set of keywords. Here, you can click the "Repeat" button, and you will see multiple messages for generating images. Wait for the robot to draw, and then find a few images you like to select.

Advanced Operation Discord Operation Command

(Here are only some parameter commands that we will use, and other parameters have little significance for us to learn AI painting.)

/Ask to obtain the answer to the question. You can ask some questions for Midjournal to answer, similar to FAQs;

/Blend image fusion, a total of 6 images can be uploaded, and sending them to the robot will help you fuse the uploaded images together to generate a new set of images;

/Docs can be used on the official Midjournal Discord server to quickly generate links to topics covered in the user guide;

/Switch to fast mode, and there is usually no need to switch this command for the duration of Fast usage;

/Help displays useful basic information and tips about the Midjournal Bot, with the help center meaning literally;

/Image uses prompts to generate an image, which is the command for generating images. Enter keywords to send;

/View information about your account and any queued or running work, including the remaining drawing duration of the account;

/Stealth for users of professional programs ($60/month): Switch to stealth mode. The meaning is that the images you generate are not displayed in the community;

/For users of professional programs ($60/month), switch to public mode, literally;

/Subscribe generates personal links for the user's account page;

/Settings to view and adjust the settings of the Midjournal Bot;

/Create or manage a custom option set;

/Preference option list to view your current custom options;

/Specify a suffix to be added to the end of each prompt;

/Show to use the image job ID to regenerate the job within Discord;

/Switch to relax mode. This mode is slower than Fast, and usually paid students will automatically switch to Relax after using Fast;

/Prefer remix turns mixing mode on/off.

/Image keyword mapping；

Enter "/", select "image", and this state will appear:

After Prompt, enter the English keywords in the format mentioned above. A set of keywords+,+spaces+keywords:

Just click on 'send' and wait for the image to appear. This operation has been demonstrated above. If you are not clear about it, please review the operation process again.

The structure and composition of keywords

Basic keywords: A basic prompt can be as simple as a word, phrase, or emoji.

Advanced keywords: can include one or more image links, multiple text phrases or words, and one or more suffix parameters.

The grammatical order of keywords

Theme: People, animals, people, places, objects, etc.

Media: photos, paintings, illustrations, sculptures, graffiti, tapestries, etc.

Environment: indoor, outdoor, on the moon, Narnia, underwater, Emerald City, etc.

Lighting: soft, ambient, cloudy, neon, studio lights, etc

Colors: vibrant, soft, bright, monochrome, colorful, black and white, soft, etc.

Emotions: composed, calm, noisy, energetic, etc.

Composition: Portrait, headshot, close-up, bird's-eye view, etc.

Example: a woman wearing street, Vintage 90's time style, interrupted 7/11 interval, neon lights, sci fi colors, noise, close up.

/Blending multiple images

Enter "/", select "blend", and this state will appear:

Select the uploaded image. The default is to upload two images, which can be added by clicking. The maximum number of images is 6:

Send and wait for the image to be produced, and then complete the task:

Image Generation Image

Upload reference images, click on the plus sign on the left side of the input box, and choose to upload images (multiple images can be selected for uploading):

Let's take a look at the official demonstration case:

Multi Prompts setting weights

You can have the Midjournal Bot consider two or more separate concepts as separators. Separating prompts allows you to assign relative importance to each part of the prompt.

Fundamentals of Multiple Prompts

Add a double colon to the prompt to indicate to the Midjournal Bot that it should consider each part of the prompt separately. In the following example, for the prompt, all words of hot dog are considered together, and the Midjournal Bot generates an image of a delicious hot dog. If the prompt is divided into two parts, hot:: dog will separate the two concepts and create a warm image of the dog.

There is no space between double colons: Multi props work with Model Versions 1, 2, 3, 4Any parameters are still added at the end of the prompt. Niji

Prompt weight

When using a double colon:: to divide the prompt into different parts, you can immediately add a number after the double colon to assign the relative importance of that part of the prompt.

In the following example, the prompt hot:: dog generated a hot dog. Change the prompt to make the word "hot" hot:: 2 dog twice as important as the word "dog", resulting in an image of a very hot dog!

[Model version] 1, 2, 3 only accept integers as weights [Model version] 4 can accept weights with non specified decimal places and default to 1.

Negative prompt weight

Negative weights can be added to the prompt to remove unnecessary elements, and the sum of all weights must be a positive number.

Image keyword suffix parameter

Aspect ratios

--Aspect, or - ar to adjust the scale of the image; Aspect ratio is the aspect ratio of an image. It is usually represented as two numbers separated by colons, such as 7:4 or 4:3.

A square image has equal width and height, described as a 1:1 aspect ratio. The image can be 1000px × 1000px, or 1500px × 1500px, the aspect ratio is still 1:1. The ratio of the computer screen may be 16:10. The width is 1.6 times the height. So the image can be 1600px × 1000px, 4000px × 2000px, 320px, 200px, etc.

The default aspect ratio is 1:1-- The aspect must use integers. Use 139:100 instead of 1.39:1. The aspect ratio affects the shape and composition of the generated image. When zooming in, some aspect ratios may change slightly.

Maximum aspect ratio

Different versions of Midjournal models have different maximum aspect ratios.

The - ar parameter will accept any aspect ratio from 1:1 (square) to the maximum aspect ratio of each model. However, during the image generation or enlargement process, the final output may be slightly modified. Example: Prompt to create an image with a 7:4 (1.75) aspect ratio using - ar 16:9 (1.78).

Aspect ratios greater than 2:1 are experimental and may produce unpredictable results

Example prompt: image/prompt verb California poppies -- ar 5:4

Common aspect ratios

--Aspect 1:1 default aspect ratio-- Aspect 5:4 Common frameworks and printing ratios-- Common in aspect 3:2 printing photography-- Aspect 7:4 Close to the High-definition television screen and smart phone screen.

How to change aspect ratio

Add -- aspect<value>:<value>, or -- ar<value>:<value>to the end of the prompt.

Chaos

--Chaos<number 0-100>Changes the diversity of results. Higher values will generate more unusual and unexpected generations;

The parameter affects the degree of change in the initial image grid - Chaos. High - Chaos values will produce more unusual and unexpected results and combinations. Lower - Chaos values have more reliable and repeatable results.

--Chaos range values range from 0 to 100. The default - Chaos value is 0.

The Effect of Chaos Generation

Low value - Chaos

Using a lower - Chaos value or not specifying a value will generate a slightly different initial image grid each time the job is run.

Example prompt: image/prompt watermelon owl hybrid

High value - Chaos

--Each time Chaos runs a job, using higher values will result in more changes and unexpected initial image grids.

Example prompt: image/prompt watermelon owl hybrid -- c 50

Very high value - Chaos

--Every time Chaos runs a job, using extremely high values will generate different initial image grids and have unexpected composition or artistic media.

Example prompt: image/prompt watermelon owl hybrid -- c 100

How to change the Chaos value

Add either - chaos<value>or - c<value>to the end of the prompt.

Negative Reminder (no)

--No negative prompt, adding - no at the end of the prompt word can prevent certain content from appearing in the screen;

Generation quality

--Quality<. 25,. 5, 1, or 2>, or -- q<. 25,. 5, 1, or 2>How much rendering quality time do you need to spend. The default value is 1. The higher the value, the higher the rendering time, and the lower the value, the lower the rendering time. A higher value also means that each job uses more GPU minutes. The quality setting does not affect the resolution.

The default - quality value is 1. Higher values will use more subscribed GPU time-- Quality accepts the following values:. 25,. 5, and 1 for the default model. The larger value will be rounded down to 1-- Quality only affects the initial image generation-- Quality applies to model versions 1, 2, 3, 4, 5, and niji.

Quality settings

Higher quality settings are not always better. Sometimes lower quality settings can yield better results depending on the image you are trying to create. The lower quality setting may be the most suitable for gesture abstraction appearances. A higher quality value can improve the appearance of architectural images that benefit from many details. Select the setting that best matches the image type you want to create.

Example prompt:/image prompt woodcut birch forest -- q.25

Using the -- quality or -- q parameter

Add -- quality<value>or -- q<value>to the end of the prompt.

Seed

--Seed<integer between 0-4294967295>Students who have used Midjournal will find that after sending prompt words, there is a very blurry noise cluster in the initial image of MJ, which gradually becomes specific and clear. The starting point of this noise cluster is "Seed". The seed number is randomly generated for each image, but can be specified using the - seed or - sameseed parameters. Using the same seed number and prompt will result in a similar ending image.

Change the seed number, use the -- seed or -- samseed parameters, and add -- seed<value>or -- samseed<value>to the end of the prompt.

Stop

--Use the stop parameter to complete the task midway through the process. Stopping the task at an earlier percentage will result in more ambiguous and less detailed results.

Using the -- stop parameter, add -- stop<value>to the end of the prompt.

Tile

--The tile parameter generates images that can be used as repeated collages to create seamless patterns of fabrics, wallpapers, and textures-- Tile is applicable to model versions 1, 2, 3, and 5- tiles only generate one tile. Use a pattern making tool like this seamless pattern checker to view collage duplicates. Version: Use the -- version or -- v parameter or use the/settings command and select the model version to use another model. Different models excel in processing different types of images.

Add -- tile to the end of the prompt.

Style

--Style<4a, 4b, or 4c>Switch between versions of Midjournal model version 4

Stylize

--The stylize<number>or -- s<number>parameter affects the strength of Midjournal's default aesthetic style applied to Jobs.

Using the style parameter, add either -- style<value>or -- s<value>to the end of the prompt.

Use the settings command to type and select your preferred stylization value from the menu.

Spotlight

--When selecting the U button for uplight, use an alternative "lightweight" upconverter. The results are closer to the original grid image. The enlarged image has fewer details and is smoother.

Upbeta

--When selecting the U button for upbeta, use an alternative beta upconverter. The results are closer to the original grid image. The enlarged image adds significantly fewer details.

Video

Use this -- video parameter to create a short video of the initial image grid being generated. Use the envelope ⊙: emoticon to respond to the completed work and have the Midjournal Bot send a video link to your direct message.

--Video is only suitable for image grids and not for high-end applications-- Video is suitable for model versions 1, 2, 3, test, and testp.

Video example:

How to obtain video links

Example prompt:/image prompt Debrent California Poppies -- video

1. Add - video to the end of the prompt;

2. After completing the assignment, click Add Reaction;

3. Select the envelope emoji;

Add -- video to the end of the prompt.

Ai Drawing Case Sharing

Rabbit from extraterrestrial space

Keywords: photo of a rabbit monster in the space, style of laurie greesley, studio ghibli, akira toriyama, James gillard, genesis impact, trending pixiv fan box, acrylic tablet knit, 4k, vibrant colors, devinart, trending on artstation, low details

Aerial island

Keywords: A communist city, built on a mechanistic flying island at the bottom, viewed from below and from afar, with soft color palettes, cinematic style, cinematography, Sharp, Hasselle, dramatic lighting, depth of field, mid shot, soft tones, 80mm, incredible high detail, Lightroom Gallery

Outer Space Astronauts

Keywords: Tiny, cute, small, astronaut in space, background is a distant planet, mini, perspective, orthographic view, movie lighting, Gediminas Prankevi č ius, Giuseppe Arcimboldo, Fujida Wulang, complex, Hyp er detailed, 4k