Midjourney’s AI Images Get Weirder And Wider

Generative AI art service Midjourney has added two new features that make images “weirder” and more panoramic.

Midjourney has been rapidly spitting out features over the past few weeks, with the service expected to make the jump to version 6 before the end of July. In the meantime, customers have been given a selection of creative new toys to play with in version 5.2.

Panoramic Images

The latest Midjourney feature is panoramic images. When a regular image is upscaled to full resolution, the user gets the option to pan the image left, right, up or down. This effectively extends the image in the chosen direction, with the AI filling in the empty space.

For example, I asked Midjourney to create a 16:9 image of a car driving down a highway, as shown below:

I then clicked the option to pan the image to the right, resulting in the following image:

Images can continue to be extended beyond one iteration, although the wider and wider aspect ratio becomes problematic.

Users can also enter Midjourney’s Remix mode and change the prompt at the same time as extending the image, effectively letting them create panoramic photo stories.

Midjourney’s pan feature is a rapid response to the Firefly AI tools that have recently been added to the beta of Adobe’s Photoshop software. Among many things, these let you extend the canvas, with the generative AI filling in the blank space.

Weird AI

The second new feature adds an element of weirdness to images.

The weird parameter can be added to image prompts, with the level of “weirdness” set between 0 and 3,000. The higher the number, the stranger the images become.

The following image of “hungry schoolchildren” had the weirdness turned up all the way to 3,000:

The weird parameter can also be used with other styles. Here, for example, I asked for a boy surrounded by aliens in the style of Stanley Kubrick, with both the stylize and weird parameters set to 700 (Midjourney recommends that both numbers are equal when creating parameter combos):

It’s debatable whether the above still looks like it was taken straight from a Stanley Kubrick movie. What isn’t in doubt, however, is that Midjourney is trying to push creators to experiment with different types of images and prompts, to generate pictures that have a distinct look, rather than adding to the pile of AI-generated mush.

Both features are available now to Midjourney subscribers.