Please make sure you are familiar with the content guidelines and are aware of what is public content in the community. You can find more detail in Content and Moderation. This document does not cover everything in the official Midjourney Discord server (#rules, #faq, #announcements, #status channels), so make sure you check it out along with the other pages in this documentation, such as Frequently Asked Questions and Resource Links.
That said, enjoy creating beautiful images! In this page:
- Basic Commands in Bot Channels
- Parameters to "/imagine"
- Stylize Values
- Quality Values
- Emoji Reactions to Generation Output
- Image Prompting with URL
- Advanced Text Weights
- Prompt Preferences and Settings
- Deprecated: May Want To Avoid
Commands are functions of the Midjourney bot that can be typed in any bot channel or thread under a bot channel. A bot channel is a channel under the "Image Generation" section on the Discord server.
/imagine creates an image based on the prompt text you provided. It produces a grid of 4 images, taking around 50 seconds with default settings.
/help displays universally helpful information and tips about the Midjourney bot.
/info shows information about your profile, plan, usage, and currently running jobs.
/subscribe creates a unique link to the subscription page of your current Discord account, without needing to sign in on the website.
/fast and /relax toggles between "fast" and "relax" mode. In fast mode, if you are out of jobs, your jobs will be incrementally billed. In relax mode, your jobs do not cost, but take longer to generate. Only active Standard and Corporate plans have access to relax mode.
/show <jobid> after recovering the ID of a job in your gallery, you can summon that job again in the chat, producing the resulting image and upscale+variation buttons. This allows you to essentially revive any job you generated yourself, bringing it into any bot channel to generate upscales and variations there, even if you have lost access to the original message.
/private and /public toggle between "private" and "public" mode. In private mode, your jobs are only visible to you. In public mode, your jobs are visible to everyone in the gallery, even if you are creating them in a thread or a DM. Access to private mode costs an extra 20$ per month.
You can find more documentation on using these in our FAQs.
Parameters are "inputs" to the command. Some are required, like a prompt, others are optional and will change how the prompt is interpreted and the image is created. They're sometimes also called "switches" or "flags".
For instance, a full imagine command might contain several things, like an image URL, some weights, and other parameters:
/imagine prompt: http://myimageonline.jpg A forest spirit at night --iw 0.2 --no trees --hd
Below are some of the parameters you can add at the end of the command, using the "--" delimiter.
{% hint style="warning" %} All parameters should always be added at the end of your command. Failure to do so might make your jobs partially or completely unusable. {% endhint %}
--beta Uses a new an experimental algorithm. This model takes significantly longer to run and only produces 2 outputs (1 for non square resolutions). It also does not work with --hd, --stylize, and --q 5.
--hd Uses a different, older algorithm that’s potentially better for larger images, but with less consistent composition. Best for abstract and landscape prompts. This also generates images at higher resolutions without the need to upscale.
--aspect, or --ar Generates images with the desired aspect ratio. Try --ar 16:9 for example, to get a 16:9 aspect ratio (~448x256).
--w <number> and --h <number> Sets width and height of the image, respectively. --ar is better supported and should be used instead. The values used by --h and --w should be between 256 and 2034, while keeping in mind that the maximum upscale resolution is around 3 Megapixels. These values work better as multiples of 64 (or 128 for --hd).
Size shortcuts for common resolutions
--ar is better supported and should be used instead
These are synonyms for longer commands. In other words, using --wallpaper is the same as saying --w 1920 --h 1024 --hd.
--wallpaper: --w 1920 --h 1024 --hd
--sl: --w 320 --h 256
--ml: --w 448 --h 320
--ll: --w 768 --h 512 --hd
--sp: --w 256 --h 320
--mp: --w 320 --h 448
--lp: --w 512 --h 768 --hd
You can read more about image sizes and how to interpret them on Understanding Image Size.
--seed Sets the seed, which can sometimes help keep things more steady and reproducible when trying to generate a similar prompt again. It must be a positive integer (any whole number between 0 and 4294967295). If not used, a random seed will be chosen instead, behind the scenes. You can react with ✉️ to a job's message to know what seed value was used.
--sameseed Makes it so the seed affects all images of the resulting grid in the same way. If not used, each image in the grid will use a different "slice" of the seed, providing higher variety.
--no Negative prompting (--no plants would try to remove plants). Equivalent to using an advanced text weight of ::-0.5
--iw Sets image prompt weight relative to text weight. Default value is --iw 0.25
--stylize <number>, or --s <number> The stylize argument sets how strong of a 'stylization' your images have, the higher you set it, the more opinionated it will be. Default value is 2500. See below for more info.
--quality <number> , or --q <number> How much rendering quality time you want to spend. Default value is 1. Higher values cost more and lower values cost less, see below.
--chaos <number> How much more varied, random, and different the results will be. Must be between 0-100. Higher values will favor more interesting and unusual generations in exchange for less reliable compositions.
--fast Faster images, less consistency, less expensive. You can also use --q 0.5 and --q 0.25 for a similar result, see below.
--stop Stop the generation at an earlier percentage. Must be between 10-100. This currently does not work with upscales.
--video Saves a progress video, which is sent to you in the ✉️-triggered DM (you must react with the envelope to get the video link). This currently does not work with upscales.
--version <1 or 2> or --v <1 or 2> Uses old algorithms 1 (which was formerly the "vibe" option, sometimes better for macro or textures) or 2, the last improvement. We are at 3 now, which you do not need to specify. So specify --version 2 to use the previous older model, or --version 1 for the one before.
--uplight Uses the "light" upscaler when selecting the U **** buttons. Results are then closer to the original image, with less detail added during upscale. Ideal for faces and smooth surfaces.
--stylize changes how artistic you want your generation to be. It can be used also with the shortened version --s, and can be followed by any integer between 625 and 60000. Here are some example values:
--stylize 625 If you basically want to turn it off and be less artistic.
--stylize 1250 Good for when you want it to be 'less strict' but still 'pretty' (this is probably recommended for skilled users).
--stylize 2500 The default value, so you don't have to specify this.
--stylize 20000 If you want it to 'take over' and start drifting from your text, but not go crazy.
--stylize 60000 Hands off the wheels, who knows what will happen. It may look nothing like your prompt.
--quality changes how much time is spent generating your image. The shortcut version is --q. Please only use one of the values below. Any other value will be rounded to a valid value instead.
--quality 0.25 Rough results, 4x faster / cheaper.
--quality 0.5 Less detailed results but 2x faster / cheaper.
--quality 1 The default value, you do not need to specify it.
--quality 2 More detailed results, but 2x slower and 2x the price (2 GPU minutes per /imagine).
--quality 5 Kind of experimental, might be more creative or detailed... also might be worse! (5 GPU minutes per /imagine).
After a job has concluded in a Discord chatroom, you will notice buttons appearing under your generation. These buttons allow you to request specific kinds of operations that are usually not available through commands alone. Please note that using these buttons to generate or upscaling images will still cost some of your limited fast GPU minutes.
- 1, 2, 3, 4 is for 🔢. Whenever you produce a grid with two or more images, you might see that some buttons also have numbers. Each number corresponds to an image, in the same order as if you were reading a comic book.
- U is for Upscale. The selected image will have its resolution increased, usually to something equivalent to 1024x1024. After upscaling an image, the following buttons might appear:
- Upscale to Max further increases the resolution of the upscaled image.
- Light Upscale Redo will redo the upscaling using a different method that tries not to add as much detail to the image. If you want this method to be your default upscale, add
--uplightat the end of your/imaginecommand. - (experimental) Beta Upscale Redo will redo the upscaling using a more efficient method that creates a much higher resolution image from the start. If you want this method to be your default upscale, add
--upbetaat the end of your/imaginecommand. Please be aware that this feature might be removed or changed substantially in the future. - (experimental) Remaster will generate variations while also using
--test --creative, very powerful but experimental parameters that can increase the quality and coherence of the image... or could also make it worse! You will generate just one image (or two with square resolutions) instead of four. Please be aware that this feature might be removed or changed substantially in the future.
- V is for Variations. The selected image will be used as a base to generate new but similar images, trying to maintain overall style and composition.
- 🔄 is for Redo. Whichever operation was carried out in the current job, it will be run again. This can produce very different results, especially when rerunning an
/imaginecommand.
"React" to a generation with various emojis to trigger actions from the bot.
✉️ :envelope: The envelope emoji reaction sends an image to your DMs with the seed # and job ID. If the result was a grid, it will send each individual image. You may have to hunt for this emoji by typing "envelope" in your emoji list.
⭐️ :star: Marks image as "favorite". This shows in a separate feed on the web gallery and sends the image to the #favorites channel.
❌ :x: Cancels or deletes a generation at any time. It is also removed from the web gallery. Please help us by removing content you accidentally generated that is in violation of our PG-13 content guidelines (see Content and Moderation).
Add one or more image URLs to your prompt and it will use those images as visual inspiration. You can mix words with images or just have images alone. For more info see Image Prompt Questions and watch the video below for a demonstration on how to upload and use an image.
--iw <value> Adjusts the weight of the image URLs vs the text. They default to 0.25. Experiment and see what you like. For example, --iw 1 makes your image URL just as important, for the generation, as your default-weight prompt text. Also see FAQ here.
{% hint style="warning" %} There is currently no way to apply different weights to different image prompts. This will be addressed in the future. {% endhint %}
Image Prompting is not the same as building on top of (or "initializing" from) a starting input image as you may see in other generation tools. Midjourney does not currently offer the ability to use a starting image, due to concerns about community public content. Instead, we let you use an image as inspiration, usually with text, to guide the generation.
You can suffix any part of the prompt with ::0.5 to give that part a weight of 0.5. If the weight is not specified, it defaults to 1. See also Text Prompt Questions.
Some examples:
/imagine hot dog::1.5 food::-1— This sends a text prompt ofhot dogwith the weight 1.5 andfoodof weight -1/imagine hot dog::1.25 animal::-0.75— Sendshot dogof weight 1.25 andanimalof negative 0.75/imagine hot dog:: food::-1 animal— Sendshot dogof weight 1,foodof weight -1 andanimalof weight 1
Prompts with a negative total weight are not allowed.
{% hint style="info" %} The "--no" command is equivalent to using a text weight of "-0.5". For instance, "--no farms" means don't include farms in the output, same as "farms::-.5" {% endhint %}
/settings — This will give you some buttons to view and change your current preferences, like the /prefer suffix command but visually. Each set of preferences is a toggle, meaning turning one on will turn the others off.
What each /settings button means
- MJ version 1 sets the
--v 1suffix - MJ version 2 sets the
--v 2suffix - MJ version 3 doesn't set any suffix: this is the default value and the current version of our tool
- Style low sets the
--s 1250suffix. If you want even less, set--s 625with/prefer suffixinstead - Style med doesn't set any suffix: this is the default value
- Style high sets the
--s 5000suffix - Style very high sets the
--s 20000suffix. If you want even more, set--s 60000with/prefer suffixinstead
- Half quality sets the
--q 0.5suffix. This is less detailed 1/2 as expensive - Base quality doesn't set any suffix: this is the default value
- High quality sets the
--q 2suffix. This is more detailed and 2x as expensive
Switching between /fast and /relax mode (more info):
- Fast mode is equivalent to using the
/fastcommand - Relax mode is equivalent to using the
/relaxcommand
- Regular upscale doesn't set any suffix: this is the default value
- Light upscale sets the
--uplightsuffix. This causes the U buttons to upscale your image while keeping the original detail mostly intact
Switching between /public and /private mode (more info):
- Public mode is equivalent to using the
/prubliccommand - Private mode is equivalent to using the
/privatecommand
/prefer suffix <text> — This will automatically append the specified suffix after all prompts you submit. Leave empty to reset.
{% hint style="info" %} Only "--" options are currently officially supported as values for prefer suffix, not just any regular text. For example, you can use "--uplight" and "--no golden hour", while "soft" or "golden hour::-0.5" will not work. {% endhint %}
/prefer auto_dm True — Jobs will be automatically DMed to you. Set False to turn this off.
/prefer option set <name> <value> — This creates a personal option, which then translates to the value you have set when you invoke it by prepending it with --. Only you can use this option. For example, /prefer option set mine --hd --ar 16:9 creates an option called "mine" that translates to --hd --ar 16:9. So you can use /imagine rubber ducks are awesome --mine, and it will be the exact same as if you did /imagine rubber ducks are awesome --hd --ar 16:9. Leave the "value" field empty to delete an option.
/prefer option list — This will list the personal options you've currently set with the /prefer option set command. You may keep a maximum of 20 personal options.
--hq --newclip --nostretch --beta --old /pixels /idea