[Advanced Tutorial] Midjourney command expansion skills

Russell

March 15th, 2023

View OriginalTranslated by Google

This article focuses on common midjourney techniques and expands the usage of some related common commands.

1.1 version switching The most commonly used is version switching, which can actually be done by adding -v followed by a number after the prompt. V4 has a better understanding of creatures, places, objects, etc., and is more accurate at picking up small details, and is also competent when dealing with complex cues with multiple characters or objects. The v4 comes in two slightly different "styles" with a slight tweak to the style of the model. Try these versions by adding --style 4a or --style 4b to the end of the V4 prompt.

Then there's Niji, an anime and illustration style collaboration between Midjourney and Spellbrush. Niji models possess extensive anime knowledge, anime styles, and anime aesthetics, making them ideal for shots with energy and action, as well as character-centric compositions of a general nature.

MJ tests are temporary releases of new models released for community testing and feedback, which can be tried occasionally. 1.2 Drawing Quality

A higher quality setting isn't always better; sometimes, depending on the image you want to create, a lower quality setting can produce better results. Lower painting quality settings may be best for representing abstract poses, while higher quality values may improve the look of images that require a lot of detail to benefit from architecture, choose the setting that best matches the type of image you want to create. half-quality = --q .5 base-quality= --q 1 high-quality= --q 2 Note that high quality is currently not supported in V4 and Niji.

1.3 Style switching

Midjourney has been trained to produce images that favor artistic color, composition and form. The --stylize or -s parameter affects the strength of this training. Low stylize values produce images that are closely related to the cues but less artistic. High stylize values produce images that are very artistic but less associated with cues.

style-low= --s 50

style-medium=--s 100

style-height= --s 250

very high style = --s750

The default value for --stylize is 100, and can accept integer values between 0-1000 when using the default [V4 model]. V4 V3 Test/Testp niji style default 100 2500 2500 NA style range 0-1000 625-60000 1250-5000 NA 1.4 Upscale Midjourney started with a low-res grid image and then attempted upscaling to increase size and add extra detail. Model Version Starting Grid Size V4 Default Upscaler Detail Upscale Light Upscale Beta Upscale Anime Upscale Max Upscale**Default ModelVersion 4512 x 5121024 x 10241024 x 10241024 x 10242048 x 20481024 x 1024 - hd 512 x 512 - 1536 x 1536* 1536 x 1536 2048 x 2048 - 1024 x 1024 v1–v3 256 x 256 - 1024 x 1024* 1024 x 1024 1024 x 1024 1024 x 1024 1664 x 1664 niji 512 x 512 1024 x 1024 1024 x 1024 1024 x 1024 2048 x 2048 1024 x 1024 - test / testp 512 x 512 - - - 2048 x 2048 1024 x 1024* - Regular (Default) Upscaler - will increase the image size while smoothing or refining details, some small elements may change between the initial grid image and the completed upscaling process.

Light Upscaler - Creates a 1,024px by 1,024px image and adds just the right amount of detail and texture. When working with older Midjourney models, the Light Upscaler can be used when dealing with faces and smooth surfaces. Detailed Upscaler is a detailed upscaler that creates a 1,024px by 1,024px image and adds lots of extra fine detail to that image. After zooming in with the Detail Upscaler, click the "Upscale to Max" button again to increase its final resolution to 1664px × 1664px. Maximum magnification is only possible in fast mode. Detail Upscaler is the default setting for Midjourney mode V1, V2, V3 and hd versions. Beta Upscaler creates a 2048 px x 2048 px image without adding too much extra detail. The beta upscaler is useful when dealing with faces and smooth surfaces. Use the -upbeta parameter to change the U1 U2 U3 U44 zoom button behavior to use the Beta Upscaler for scaling. Anime Upscaler is the default upsampling program for -niji models. It scales the image to 10.24 px by 10.24 px and is optimized to work well with illustration and anime styles. Use --upanime parameter to change U1, U2, U3, U44 upsampling button behavior to use anime upscaler for upsampling. Remaster reshaping is one of the further expansion options made by the V1, V2 or V3 model version. Remaster will re-extend the image that has been expanded through the -test and -creative parameters, mixing the original image composition with the new one. The test model has the same nature and characteristics. Any previously extended mission can be remade by clicking the Remake button found on the adjustment above the original. To recreate very old quests, use the /show command to refresh quests in Discord. 1.5 Mode The /stealth and /public commands toggle between stealth and public mode. The trial version only has an option for public mode, and Pro plan subscribers can access incognito mode to prevent their pictures from being seen by others on the Midjourney website.

Fast and Relax Modes Standard and Pro plan subscribers can create an unlimited number of images per month in Relax Mode, which requires no GPU time, but jobs will be queued for processing based on how much you use the system.

Remix mode deserves attention, it can change the aspect ratio between hints, parameters, model versions or variants. Remix incorporates the general components of a starting image into a new job. Remix can help change image settings or lighting, develop subjects, or achieve tricky compositions. Remix is an experimental feature and may be changed or removed at any time.

Enable remix mode by activating it with the /prefer remix command or using the /settings command and toggling the ️Remix Mode button. When Remix is enabled, it allows you to edit prompts on a per-variable basis. To remix a premium product, choose Make Variations.

You can add or remove parameters in Remix mode, but only with valid parameter combinations. Changing "/imagine prompt illustrated stack of pumpkins --version 3 --stylize 10000" to "illustrated stack of pumpkins --version 4 --stylize 10000" returns an error because MJ4 is not compatible with the Stylize parameter. 1.6 Custom Preferences Custom Preferences Custom options can be created using the prefer command, which automatically adds commonly used parameters to the end of the prompt. /prefer auto_dm will automatically send a private message after completing the task /prefer option option to create or manage custom options /prefer option list View the current list of custom options /prefer suffix specifies the suffix to add to the end of each prompt /prefer option set creates a custom parameter that you can use to quickly add multiple parameters to the end of the prompt. You can set up to 20 options.

PreferOptionSet /prefer option setmine --hd --ar 7:4 Creates an option named "mine", which stands for --hd --ar 7:4.

PreferOptionSet_Used Using the /imagine promptvibrant California poppies command, mine is equivalent to /imagine promptvibrant California poppies --hd --ar 7:4. Setting the value to null is equivalent to removing the option. /prefer-suffix will automatically append the specified suffix to all prompts, use the command with no value to reset. For example: /prefer suffix --uplight --video Midjourney parameter description You can add parameters to the prompt and change how the image is generated. 2.1 Basic parameters Aspect Ratios. --aspect or --ar can change the aspect ratio of the generated image. Different Midjourney versions of the model have different maximum aspect ratios

Version 4 Version 3 Test / Testp niji Maximum aspect ratio 1:2 or 2:1 5:2 or 2:5 3:2 or 2:3 1:2 or 2:1

The --ar argument will accept any aspect ratio from 1:1 (square) to the maximum aspect ratio of each model. However, the final output may be slightly modified during image generation or upscaling. For example, using the "--ar 16:9(1.78)" hint creates an image with a 7:4 (1.75) aspect ratio. --aspect 1:1 default aspect ratio --aspect 5:4 common screen and printer aspect ratios --aspect 3:2 often used in print photography --aspect 7:4 close to HD TV screens and smartphone screens Chaos:

--chaos <number 0-100> changes the degree of variance of the results, higher values produce more unusual and unexpected results and combinations, while lower --chaos values output more reliable, repeatable results. NO: --no Negative hint, use this parameter if you don't want something to appear in the generated result.

quality : It has been mentioned in the previous section, so I won’t repeat it.

Seed: --seed <integer between 0-4294967295> Anyone playing machine learning knows that using the same hint and random seed guarantees the same output (V4 and Niji only).

How can I find the corresponding seed value for the generated image? Please see the picture below.

Stop: --stop <integer between 10-100> Use the --stop parameter to stop generation at the specified scale, which will produce a softer, less detailed initial image.

--stop affects the verbosity of the final Upscaler results

Style:

--style <4a or 4b> Switch between different versions of MJ4.

Stylize:

The --stylize or --s parameter affects how much Midjourney's default aesthetic is applied to the job. This parameter has been mentioned before and will not be repeated. Tiles:

The --tile argument generates an image that can be used as a repeating pattern to create seamless patterns for fabrics, wallpapers, and textures. Only available for V1, 2 and 3 versions.

2.2 Upscaler parameters

Uplight: --uplight Use an alternative "light" upscaler when the U button is selected. The result is closer to the original grid image. The enlarged image has less detail and is smoother.

Upbeta: --upbeta Use another beta upscaler when the U button is selected. The result is closer to the original grid image. Zoomed-in images add significantly less detail.

2.3 Other parameters

Creative --creative Modify the test or testp version to be more varied and creative. Note: no need to add value

Image Weight --iw sets image hint weight relative to text weight, default is --iw 0.25

Sameseed --sameseed seed value Create a large random noise field that is applied to all images in the initial grid. When --sameseed is specified, all images in the initial grid use the same starting noise and will produce very similar generated images.

Video:

Use the --video argument to create a short video generated from the initial image grid. React to work done with the envelope ✉️ emoji and ask Midjourney Bot to send a video link in your direct message.

--video only works in V1, 2, 3, test, testp.

Detailed steps to get the video. 1. Add -Video at the end of the tip 2. Complete the task and click add reaction 3. Select ✉️ emoji (Search for envelopes in the search box)