Artificial intelligence is revolutionizing content creation. Design is undoubtedly one of the areas in which the developments are most impressive. It is now possible to create any image straight from our imagination with a disconcerting realism.

One of the most popular tools to achieve this kind of feat is Midjourney. However, for a result that perfectly coincides with what you had imagined, you need to learn to speak the language of this AI.

In this article, we will share with you the best practices for writing a good prompt for Midjourney. Speaking of prompts, we have already published an article to learn how to make good prompts for AIs. We invite you to go read it if you haven't already.

About Midjourney

We are not going to dwell on the presentation of Midjourney. Nevertheless, it is important for you to know that the AI has a default image style.

If you do a very short prompt for Midjourney with a single word, like "lion", you will get proposals like below.

lion images midjourney

This definition that strongly resembles some Instagram filters is the default for Midjourney. Nevertheless, to get a unique result and avoid your visual being just another AI edition, you should create ultra-descriptive prompts.

When writing your prompts, think carefully about all the details. Anything you don't talk about will be randomly defined by Midjourney.

Write your prompts specifically for Midjourney

Each AI responds differently to the prompts submitted to it. Midjourney, is not as sensitive as humans to writing. Here is what you need to consider when writing a prompt for Midjourney.

The punctuation

Most of the punctuation you will use in your prompts will not be taken into account by Midjourney. Nevertheless, we recommend that you keep punctuation (especially commas and periods) to organize your own thinking.

This being said, the only punctuation that Midjourney understands is the following:

  • Double hyphens delimit parameters, for example "--ar 16:9" or "--v 5".
  • Double semicolons delimit integers (positive and negative) for weights. For example, "::-0.5".
  • Spaces are used to separate each query. This concept is not very important in most cases.

Exclusion

Midjourney does not understand exclusion. Here is the result when you ask Midjourney for "a bicycle without wheels"

bicycle images midjourney

It is very difficult to remove an element with Midjourney. Nevertheless, you can try to solve it by adding a "--no {{element to remove}}" instruction.

Giving structure to your prompts for Midjourney

To make sure that your prompts are all structured with the same base and therefore easy to modify, here is a structure that you should follow.

1. The image style

In the first part of your prompt, fill in everything about the definition of your image. If you don't do anything, your image will look like yet another Midjourney generation, like the lion above.

You can ask Midjourney to imitate the style of a painting or a cartoon by suggesting artists to base it on. Here is an example when asked: "Van Gogh style painting of a lion".

lion images van gogh style midjourney

However, where Midjourney is really interesting is in its ability to reproduce images that really look like photographs.

To do this, there are several points to consider.

  • You can choose the type of camera that the AI should imitate: SLR for film cameras and DSLR for digital SLRs.
  • The type of lens can also be proposed (18, 35, 50 or even 250 mm).
  • The model of the camera.
  • The recommendations : they are preceded by "--" and allow to guide the AI on certain aspects concerning the version of the AI (v 4, v 5), the image format (3:2, 16:9, etc) or the style parameters (s 100 for medium, s 250 for high or s 750 for very high)

Here is what Midjourney generates when asked "Photo with a Canon R5, 50 mm, DSLR, of a lion --v 5 --ar 3:2 --s 750 - no camera".

A little aside that is important: "--no camera" avoids Midjourney to insert a Canon R5 in the picture.

lion images canon r5 midjourney
Isn't it amazing?

Once you have found the resolution that suits you to generate your image, you have done the hardest part. What's left simply requires you to let your imagination run wild.

2. The Subject

In the example just before, our subject was the lion. Except that sometimes you don't have just one subject, but several. If this is your case, take care to describe each one well.

When necessary, list the number of individuals. For example, instead of saying "a group of lions" say "a pack of 3 lions".

This is what Midjourney generates when submitted the following prompt: "Photo with a Canon R5, 50mm, DSLR, pack of 3 lions --v 5 --ar 3:2 --s 750 --no camera"

group of lions images midjourney

3. The environment

To give context to your image, put your subjects in an environment. Midjourney allows you to make unthinkable life scenes ultra-realistic. Let's imagine that our three lions are walking through the streets of San Francisco

We could then propose a prompt to Midjourney such as "Photo with a Canon R5, 50 mm, DSLR, pack of 3 lions walking through the streets of San Francisco --v 5 --ar 3:2 --s 750 --no camera"

group of lions images midjourney

Admittedly, the AI sometimes has trouble counting to 3. But let's face it, the result is amazing.

4. The lighting

Depending on the time of day, the lighting is not the same. Thus the colors and contrasts differ. Here again, Midjourney is able to reproduce these contrasts.

Let's ask Midjourney to generate a photo of a lion in San Francisco by adding the word "golden hour". The prompt would look like this: "Photo with a Canon R5, 50 mm, DSLR, a lion walking in the streets of San Francisco, golden hour --v 5 --ar 3:2 --s 750 --no camera"

Here is what Midjourney generated:

group of lions images midjourney

5. The angle of view

Another factor you can play with is the viewing angle of the image. In some cases, you will need to adjust your optics to make this possible. But it can be a real game changer depending on what you want to portray.

Here are some examples of angles of view you could incorporate into your prompts for Midjourney:

  • For distant shots: "Wide-Angle Shot", "Ultra-Wide Angle", "Far-Shot", "Aerial View" or "Satellite View"
  • For medium shots: "Medium-Shot", "Top-Down Shot", "Ground-Shot", "Ground-Shot", "Low-Angle Shot", "High-Angle Shot", "Full-Shot", "Full-Body Shot" or "Full-Length Shot"
  • For shots very close to the subject: "Eye-Level Shot", "Close-Up", "Glamour Shot", "Glamour Portrait", "Macro-Shot" or "Macrophotography"

Here is what we get with the following prompt: "aerial view with a Canon R5, 50 mm, DSLR, streets of San Francisco, golden hour, --v 5 --ar 3:2 --s 750 --no camera"

 streets of San Francisco golden hour images midjourney

Finally, by taking care to follow each of the steps listed above, you can generate any image. Good prompts for Midjourney are prompts that allow you to get the image, resolution, details, angle of view and lights you want.

Depending on your needs and desires, you can even create your own universe by giving a maximum of information about the colors, the atmosphere or the shapes to adopt.

If you combine Midjourney's visual creation with Abyssale's ability to scale and automate, you'll push the sliders to a level never before reached in terms of visual creation.