Inside Midjourney, the generative art AI that rivals DALL-E

Midjourney-generated imagery showing yellow Minions against a pastoral backdrop (left) and an abandoned office building with large windows (right)

Images generated by Midjourney

AI tools like DALL-E that generate images from short text descriptions are all the rage these days, so it’s only logical that other systems would start competing for space in the burgeoning AI art market.

Last week, the company behind the generative AI art tool Midjourney opened its closed beta, which allows anyone to sign up for an account and create surreal renderings – albeit with a limited number of “free” generations. The influx of new users was a stress test for the system, which rivals DALL-E in its ability to produce stylistic and sometimes hauntingly realistic renderings.

“In the next 24 hours we will try to open beta access to everyone,” Midjourney CEO David Holz wrote in a message posted to the AI ​​tool’s Discord server on Monday night. “We don’t know if our servers are ready, but we’re ready to find out.”

Unlike DALL-E, Midjourney seems particularly good at creating environments, particularly fantasy and dystopian sci-fi scenes with dramatic lighting that look like rendered concept art from a video game.

Image generated by Midjourney showing the prompt:

Image generated by Midjourney with prompt: “Interior of abandoned office building where secret raves were thrown”

Images generated by Midjourney with prompt: “Spaceship interior full of lush plants”

It’s also pretty good at making surreal pastiches that mimic different art styles. Here are some examples where I’ve tried to recreate classic art featuring celebrities like Megan Rapione and Minions:

Images generated by Midjourney with the prompt:

Images generated by Midjourney with prompt: “Megan Rapinoe piloting a Gundam in German Impressionist style”

Image generated by Midjourney showing the prompt:

Image generated by Midjourney with prompt: “Photorealistic Minions on Windows XP Background”

The test gave way to a fully open beta on Wednesday, allowing anyone to sign up and join the project’s Discord channel. The beta runs entirely through Discord, with users typing their prompts directly into the chat interface and receiving messages from a bot showing their generations rendering in real-time. Users can then choose to upscale and enhance an image from each set of generations, or create more variations from the same prompt.

However, this “free trial period” only gives each user a limited number of generations before the bot prompts them to purchase a subscription. For non-commercial use, the cheapest plan allows 200 images for $10 per month, while the premium tier allows for unlimited generations for $30 per month. (The makers of DALL-E also plan to offer paid access to their AI tool, and recently polled their beta testers about possible price points.)

Midjourney also — thankfully — discourages people from coining NFTs by requiring anyone who generates images into “anything to do with blockchain technologies” to pay a 20% royalty on any sales over $20,000 US dollars per month.

However, it’s still not clear how the hype surrounding generative AI art will translate into a business model – especially when the large AI models on which they are based have been repeatedly shown to replicate racist and sexist stereotypes.

While the makers of DALL-E have tried to mitigate some of the training biases inherent in these models, Midjourney hasn’t released any information on what datasets and methods were used to train its AI tool, and there doesn’t seem to be much explicit content protection apart from the automatic blocking of certain keywords.

The “Content and Moderation” section of the Midjourny User Guide instructs users not to “create images or use text prompts that are inherently disrespectful, aggressive, or otherwise abusive” and “create visually shocking or disturbing content,” including Adult Content and Gore . The rules also ban content that “may be viewed as racist, homophobic, disturbing, or in any way derogatory to a community,” including “offensive images of celebrities or public figures.”

It’s unclear how or how well any of this will be enforced, but with results this impressive, the Midjourney project will likely be something to watch as companies begin to chart a precarious path forward for imaging AI.

Leave a Comment

Your email address will not be published.