MIDJOURNEY TUTORIAL: INTRODUCTION

This tutorial guides you through an experience of AI image generation without encouraging you to generate your own images. It does this to bring your attention to the ethical dilemmas and aesthetic issues involved in image generation before you engage in the practice.

It invites you to refrain from contributing to the overproduction of images through free labor performed for AI companies until you have a better understanding of what AI image generation is and can make a more informed decision about engagement. This is a consentful tech approach to AI image generation.

What is Consentful Tech?

'Consent' means to give permission for something to happen. You may, for example, not want to consent to giving your personal information to a company in order to access its app. In this case, you may not want to consent to contributing free labor to AI companies to train their machine models in order to learn how image generation works by testing it out. As creators of the Building Consentful Tech zine Una Lee and Dann Toliver write, "Consentful technology is about having control over our digital bodies."

What is Midjourney?

The Midjourney machine learning model is a text-to-image and image-to-image model created by the Midjourney research team led by tech entrepreneur David Holz. It was founded in 2021 and released to the public in 2022. It allows users to engage in AI image generation with text prompts and other images.

What is an AI Prompt and How Does it Work in Midjourney?

An AI prompt is an input given to an AI model to guide it in generating a particular outcome. The Midjourney team has not released its model's code and weights (note: these terms are defined in the tutorial), so exactly how its model works is unclear.However, since Midjourney used an open source version of the code of another model -- Stable Diffusion, created by the company Stability AI -- there is a general idea of its process (note: this is also discussed in the tutorial). The Stable Diffusion source code can be downloaded for free, redistributed and modified under a permissive license known as Creative ML OpenRAIL-M.

Why Did Midjourney Allow Users to Make Free Images with its Model?

Midjourney used to give users 25 free images before requiring a subscription. It stopped doing this in 2024, after this tutorial was created. Midjourney allowed users to generate free images because the more that people use Midjourney, the better the model is trained. Also, when people upload images to Midjourney -- something this tutorial does not encourage you to do -- the model gets access to images that the Stable Diffusion model, which its research team used to create it, may not have used in its initial training.Stable diffusion was trained on LAION-5B, a controversial dataset of over 5 billion text and image pairs. The images in the dataset were scraped off the Internet without consent from copyright holders and other individuals like artists who made their images available online without knowing that AI companies would be able to scrape them and use them to train AI models.

Explore Midjourney. Do Not Subscribe.

Visit the Midjourney Website. Go to the 'documentation' area. Click on 'Getting Started'.Go to the left menu bar and scroll down to 'Writing Prompts'. Click it and select 'Prompts'. Read through the definition of a prompt.

What is a token? How does the Midjourney website describe the way that a prompt works?

Scroll down to the section 'Advanced Prompts'. Look at the image of a Midjourney prompt.

Describe the four parts of its syntax (Note: the word 'syntax' refers to the way words are put together to form phrases, clauses and sentences).
The section of the prompt that reads '/imagine' is called a slash command. Why do you think it is called that? Start a list of the slash commands that you encounter over the course of the tutorial. The command '/imagine' should be the first on your list. Write the word 'prompt' next to it to describe what it does.
What kind of data goes in the 'image prompts' area of the prompt?
What do you think you would type in the 'prompt' area?
What is a 'parameter'? What do you think that the 'parameters' part of the prompt might do to the image that you generate?
Notice the order of the various elements. Write it down so that you remember what comes first, second and third in an /imagine slash command.