How to write prompts in Midjourney

If you’re reading this article, most likely you want to get the most specific output from the generated images.

The answer is simple - use parameters to make your prompt as detailed as possible. Midjourney gives the ability to change the quality, aspect ratio, tile of the image and even remove unwanted parts from it.

Let’s learn how to do this in 2 easy steps!

Generate and enhance the image

/imagine

To generate an image, you’ll need to write a prompt. To do this, type /imagine – it will open a box where you can type.

Midjourney will give you four generated images with U and V buttons. Here, U means upscale, and you can pick one of four images and upscale it.

Although you can upscale the resolution, you can’t change the DPI amount in Midjourney. To adjust your DPI, you can use any graphics editor(like Adobe Photoshop, Topaz Gigapixel or others).

If you want to tweak the image and get something similar, use V button with the image number, which stands for image variation.  

That will open the remix prompt, where you can specify what you want to change, add or remove specific details or change the style, aspect ratio, etc.

If you like an image and want it to be a style reference for future generations, you can download it and add it to the prompt or get its seed number.

/describe

Additionally, instead of a prompt you can add a picture to it to create something similar to the image you’ve uploaded. In these cases, the /describe command can be used to describe pre-existing images similar to the result you want to achieve and understand how Midjourney interprets them.

Type the command, and a drop and drag box will open. Here you can upload the image and generate results based on it.  

Enhance the output

Also, you can customize the final output by using commands that appear after you generated the images.  

Vary (Strong) and Vary (Subtle) generate a new grid of four images based on the output.

Zoom Out 2x, Zoom Out 1.5x and Custom Zoom will extend the canvas without changing the content of the original image. The new area will include the information given in the prompt.

The pan buttons ←→↑↓  will expand the canvas in a chosen direction without changing the content. The new area will also include the information given in the prompt.

Midjourney Parameters

The most important part of generating in Midjourney is the parameters.

Parameters are optional commands that change the final image. Midjourney accepts more than one parameter. Add them in the end of the prompt by typing--, the parameter name and specifics. 

Style

--style or--style raw

To get more specific results, give Midjourney a prompt specifying the style you want to achieve. Here you can go creative and mention the style itself, artist, genre, artform or anything else for your output.

You can even ask Midjourney to create a watercolor sketch to get the children’s book illustration output. As for the artist, you can use not just painters' names, but cinema directors, writers, designers or anyone else.

Style raw is a parameter for advanced users who needs no style influence, which Midjourney usually adds by default.

To give your prompt more immersive colors, you can also mention the rendering/lightning you expect to see, like cinematic or volumetric lightning, octane render, long exposure, etc. You can also specify the camera lenses, so Midjourney will give you even more specific results.

Quality

--q,--quality(and the number .25, .5 or 1)

Of course, don’t forget about the resolution - this parameter gives better images but generates the image longer, taking more GPU minutes to do the job. The default value is 1 and larger values are rounded to 1.

For example, here are two quality versions, lower (.25) and default (1):

Aspect Ratio

--aspect,--ar(from 1:1 to 7:4)

This parameter changes the output aspect ratio. The image could be 1000px × 1000px, or 1500px × 1500px, and the aspect ratio would still be 1:1, which is the default ratio.

--aspect 1:1 Default aspect ratio.
--aspect 5:4 Common frame and print ratio.
--aspect 3:2 Common in print photography.
--aspect 7:4 Close to HD TV screens and smartphone screens.

Here’s the result with different aspect ratios:

Stop or No commands

--stop(and number from 0 to 100)

This parameter stops the rendering of an image halfway and can provide an incomplete output for various needs or to save GPU render minutes.

--no(item1, item2, item3)

Removes the specific element from the image. Pay attention, that Midjourney won’t understand commands like don’t add, without or but. For example, you can write the prompt and specify that you want --no fruit

Stylize

--stylize,--s(and number from 0-1000)

Use this command to stylize the output more or less - pick a number from 0 to 1000 to affect the final result.

Here are the two stylized versions with different outputs:

Seed

–-seed ( and the number from 0 to 4294967295)

The seed number is used to create a field of visual noise, like television static, as a starting point to generate the initial image grids. Seed numbers are generated randomly for each image but can be specified with the seed parameter.

If you use the same seed number and prompt, you will get similar final images. 

Tile

--tile

Is a great command to generate repeating patterns, wallpapers and textures for any case scenario.

Here’s generated tile example:

Chaos

--chaos(and number from 0-100)

Another interesting parameter you may use, the chaos increases the abstraction level and randomness in the output.

Here you can take the number from 0 to 100 to variate the result:

Weird

--weird,--w(and number from 0-3000)

A new experimental feature that will allow you to add something unexpected to your prompt. The default number is 0.  Try starting with smaller values, such as 250 or 500, and then go up/down from there.

For example, here’s a weird parameter at 0 and 1000:

To sum up

Midjourney is a great tool to operate for your creative needs. With the parameters above, you can get and vary an amazing output, making it as close to your idea as possible.

To learn more about AI tools, check out our educational design blog!