Recently, I’ve discovered a discord server named Midjourney. This is not any regular server you are used to. In fact, it is an interactive artificial intelligence application where you can generate high quality images based on text description. Much like DALL-E, but more advanced. It has several algorithmic choices where each has a different path of visualizing your input. Here are some examples I created using the tool:
Midjourney also takes any image as an input, as long as you give some description text with it. I used my own photographs as inputs and described my desired output in such format: “[img url] using this face as a reference, a boy wearing a hoodie, drinking coffee, in front of a laptop.”
With different variations of my input images and various similar contexts, the above images were generated. Midjourney first generates a blurry sketch and builds onto it step by step. When the final output has been generated, you will generally have four different images. Within the Discord server, you can choose to advance one of the four outputs or get variations of them simply by clicking on buttons provided to you.
You can rate the results with emojis, and these act as feedback to further improve the algorithms. The free trial consists of 25 jobs/requests, and the buttons I mentioned also count as one point deducted. They are still in their beta phase as a project, and I assume in the near future, we will see unbelievable improvements within this project compared to today.
