Midjourney
A popular text-to-image service from an independent San Francisco lab, launched in open beta in July 2022 and first used entirely through a Discord bot.
Also Midjourney bot · /imagine
Midjourney, Inc. (David Holz), 2022
You type a description into a chat, and a few seconds later you get four painted or photo-like images to choose from.
Where it fits
+All relationships as a list
How it works
Midjourney is a small, self-funded lab founded by David Holz (previously co-founder of Leap Motion). Its image generator entered open beta on July 12, 2022, and was used through a Discord bot: type /imagine followed by a description, get a grid of four images, then upscale or vary your favorite.
Unlike Stable Diffusion, Midjourney is closed: its architecture and training data are not published. It is generally assumed to use diffusion, like other text-to-image systems.
It became known for a polished, painterly default look, and its images frequently went viral, including fake but convincing news photos. That made it a central example in debates on deepfakes and on training generative models on artists’ work without permission.
Everyday use cases
- Case 01
How you've probably met it
Through viral AI art in 2022 and 2023, such as the fake photo of the Pope in a white puffer jacket, and in concept art and mood boards.
Strengths, weaknesses and when to use it
Strengths
- Strong artistic style and image quality with short prompts.
Weaknesses
- Closed model with undisclosed training data; subject of copyright lawsuits from artists and studios.
When to use itAs an example of a widely used, closed text-to-image product.