The perfect picture with Midjourney: all important prompts and parameters at a glance

A realistic picture of a cat, a cartoon elephant, a Polaroid of two friends playing chess: Midjourney creates almost any image you want. But in addition to the description of the desired content - the so-called prompt - there is a whole range of settings, parameters and additional functions that you can use to get exactly the picture you want.

Midjourney: Basic operation of the AI image generator

Although Midjourney is working on its own web interface, image generation is currently still running via the Discord chat service. You can generate images here by giving the Midjourney bot the command /imagine and writing your prompt directly after it.

Midjourney generates four image variations for each prompt. To select one of them and receive it as a single image, simply click on U1, U2, U3 or U4. You can also use any of the four suggestions as a basis for further images. To do this, click on one of the buttons labelled V1 to V4. The style and composition of the variants generated in this way are based on the selected image.

Once you have decided on an image using one of the U buttons, Midjourney gives you a range of further options for editing the image. With Vary (Strong) and Vary (Subtle) you can create strongly or subtly altered variations of the image.

With Vary (Region), on the other hand, you can exert a much more targeted influence. To do this, select a specific area of the image and enter a new prompt for this part of the image only. This allows you to place an additional object in an image, for example.

The Upscale (2x) and Upscale (x4) buttons increase the image resolution by a factor of two and four respectively. There are also three buttons for zooming out. Midjourney generates the image areas around the original image.

The left, right, up and down arrow buttons work in a similar way. The only difference here is that the actual image is ‘shifted’ in the specified direction.

Tips for the perfect Midjourney prompt

Midjourney usually produces usable images even without overly complex prompts. Nevertheless, you should try to be as precise as possible when describing the desired image content. At the same time, you should also remember not to describe the image with an unnecessarily large number of terms.

Midjourney takes aspects such as the object to be shown, the medium, the surroundings, the lighting, colours, mood and composition into account when calculating the image. Depending on the desired result, you should take these aspects into account when creating the prompt.

A good source of inspiration for prompts is the Prompt Hero website. Here you can search for prompts and get sample images generated with Midjourney.

Working with image templates in Midjourney

Even if it may seem so at first glance: Midjourney is not only capable of text input. In fact, you can also provide the AI with images. You first have to upload them using the plus button. Then copy the image URL and paste it after the /imagine command. Then enter your text prompt and Midjourney will create a new image based on the template and text.

When combining image and text, it can be helpful to increase or decrease the weighting of one or the other. To do this, Midjourney gives you the parameter --iw, which you must place after your prompt. You can give the parameter a value between 0 and 2. The higher the value, the greater the influence of the uploaded image. If you do not use the parameter, the value is automatically set to 1.

Tip: We explain how to use image prompts to use one and the same character in a series of images here.

Examples: We created these images with Midjourney

The most important mid-journey parameters explained

We have already introduced one parameter. But there are a few others that you should definitely know about. The --ar parameter in particular will save you a lot of time in post-processing. This is because you can use it to set the aspect ratio. For example, if you put --ar 16:9 at the end of your prompt, Midjourney will generate the images in an aspect ratio of 16 to 9.

The --no parameter is very helpful for specifying the actual content of the image. You can use it to specify which things should not be visible in the image; --no dogs, for example, would ensure that no dogs are visible in your image.

Another parameter you should definitely experiment with is --stylise or --s for short. Midjourney has been trained to produce the most artistic results possible. However, this is not always desirable to the same extent. If you enter a value between 0 and 1,000 after --s, you can determine how artistic the result should be. The default value is 100, so depending on the desired result, it is worth experimenting with higher or lower values until the AI spits out a suitable image.

You can use the --chaos parameter to specify how much the four images generated by default should differ. Here you can select a value between 0 and 100. The higher the value, the more the four images will differ. The default value is 0.

The --weird parameter, on the other hand, can help to deliver rather unexpected results. Here, too, experimentation can be fun, but the Midjourney makers recommend starting with lower values for good reasons. By default, the --weird value is 0, but you can increase it to up to 3,000 by entering the parameter.

Midjourney is not intended for the creation of moving images. However, you can still create small video animations with the image AI. However, these only depict the image creation process.

To create clips, you must append the --video parameter to your prompt. Once the prompt has been executed, you must then select the selection menu for reactions to the generated message in Discord and then search for the envelope. Click on it and Midjourney will generate a video link for you.

Last but not least, we would like to introduce you to the --tile parameter. This is particularly interesting for creating backgrounds or textures. Images whose prompts have been provided with this parameter can be used as repeating tiles. In other words, the image can be strung together an infinite number of times to create a pattern.

Ordering several images at once: how permutation prompts work

Thanks to permutation prompts, you can create several different image orders at once. To do this, you need to write several terms in angle brackets and separate them with commas. Midjourney then generates a separate prompt for each term listed there and executes it.

A simple example would be the prompt /imagine a {red, green, yellow} cat. This would result in a total of three prompts. Another practical feature is that you can also apply the function to parameter values. For example, you can use the addition --s {0, 300, 500} to create and execute a prompt with three different stylise values.