Midjourney’s Site of Production
I tried my hands on Midjourney1 yesterday. The results are excellent, en par with what ChatGPT produces, even for weirder prompts. They are not without fault, but they pass. What fascinated me the most, was the process on how to obtain an image.
All other providers of text-prompt to image generators use a form, on a website. They provide their service either freely accessible or through an account. Midjourney, on the other hand, is completely Discord-based. This raises some interesting questions and observations.
First of all... why?
Realizing such a process is definitely more complicated than having a simple form on a website. The backend is probably similar then with every other provider. Piping the user’s input into the neural network and returning the image might be very different. I was wondering if that was a conscious choice, or accidentally. It might actually be, that the skill-sets and know-hows present in the Midjourney team was more on the site of how to code bots (instead of, say, web development). That would be a fascinating paradigm to dive into… not UI based development, but conversation-first.
Secondly, a collaps of the site of production and the site of audiencing on the user's side.
Having a look at Midjourney’s image generation process, through the critical analysis of visual culture, also provides some interesting observations. In Midjourney, one has to join a Discord server and talk to a bot, to produce images from text-prompts. You have to join one of the many #newbie channels, with many other users. As such, your image requests can be seen by everybody else, whereas you also see what everybody else requests. It’s quite a busy place, and often you have to scroll through the chat feed to find the answer of the bot, which contains the results of the request.
This shared space and process of image generation automatically introduces some social restrictions. Users are probably much less tempted to request images that would be socially unacceptable, like pornography or violent material. I was able to observe some users asking for idealized versions of women or men, but it was more in line with what was fitting for advertisement.
Anyway, here is what I requested… a capybara relaxing in a hot tub, surrounded by other animals and nice plants.
