Making better images using “Prompt Engineering” on Midjourney — Part 1

The role of designers is already changing and will continue to do so as AI technology advances. With the help of AI, we will be able to automate many of the repetitive, time-consuming tasks that we currently perform. This will free up more time for us to focus on creative tasks and problem-solving. AI can also assist us in making more informed decisions by providing data-driven insights and analysis. However, we will still need to maintain our unique human perspective and creativity to ensure the designs are effective and engaging for users.

Midjourney prompt: mona lisa horrified after looking at her painting — chaos 100

Having said all that, in this post, I will:

1. Discuss Midjourney and their newest version (v5);
2. Define prompt engineering;
3. Devise ways to write better prompts for Midjourney.

Midjourney is a research lab that uses AI (NOT stable diffusion, according to Midjourney) to produce images based on textual descriptions. One can generate better images by prompt engineering. Prompt engineering is an AI (Natural Language Processing, specifically) concept that involves giving clear and precise descriptions to the AI as input to get better results.

Midjourney is a powerful tool that can help us, designers, to design better. First, there are a few points one must remember while using Midjourney:

1. Midjourney has recently released version 5. Users can use the version with “ — v 5”, but they must be paid members. There are a few differences between versions 4 and 5, according to live office hours they held on Discord recently. The following are the differences:
2. All four images generated would be upscaled by default in v5.
3. Prompting should be in sentences in v5 (rather than random keywords separated by commas). Users can use chatGPT or character.ai to generate prompts.
4. Users can add — tile to the prompt to generate “tile-able” images in v5.
5. Generated images would be more photo-like in v5.

Having that out of the way. There are a few parameters that can give us more control and better (more artistic) results. These are the parameters I frequently use:

— iw <number> : image weight i.e. how much weight should be applied to the focal point of the image

— stylize <number 0–1000> : how artistic weight should be applied to the image. Read this midjourney page to get a better idea.

— no <text> : what not to include in the images.

— chaos <number 0–100> : how much variety to include in the composition of the images. Read this midjourney page to get a better idea.

NOTE: users can set multiple parameters in the same prompt. Example:

Source: https://docs.midjourney.com/docs/parameter-list

Users can add style to images by using design styles, art styles, designers, and artists in the prompt. I have compiled a list of design styles and designers and how they can be used in prompts. More about that in parts 2 and 3…