Midjourney's Pan Feature Explained (Up, Down, Left, Right)
Midjourney gave us zoom and now they give us pan!
Midjourney's latest feature, 'pan', allows users to expand their AI artwork by adding pixels to the left, right, top or bottom of an image.
The introduction of the 'pan' feature comes just after Midjourney unveiled the zoom functionality, which is a game changer for many looking to expand the creative landscapes they are working in.
Pan works differently than the zoom feature, but both currently work on any upscaled v5+ image.
Read below to learn the basics of the Midjourney 'pan' feature. Lots of example images included!
How to pan in Midjourney - the basics
Create a new, upscaled image with Midjourney v5+ or use the /show command to retrieve one of your existing v5+ upscaled images (type the job id of the upscaled image after /show).
Click one of the arrow buttons below your upscaled image to pan the image left, right, up or down
If you have Remix Mode turned on (type in /settings to check this), you will get a pop-up after you click a pan arrow. Here, you can alter your prompt, if you want to, and have greater control over what shows up in the newly panned section of your image.
View your results grid and pick an image to upscale or re-roll for new options. Note: The variations buttons (”V” buttons) are not available on panned images.
Panning adds pixels to your starting image in v5.2
Unlike the zoom out feature in Midjourney, panning actually adds pixels to your starting image. If your first image is a square 1024x1024 pixels and you repeatedly pan to the right, you will get images that are 1536x1024, then 2048x1024, then 2560x1024, etc. This only happens in v5.2, not in v6+.
What happens if you zoom out on a panned image?
Your new zoomed out images will revert back to the standard image dimensions for the version of MJ you are using, regardless of how many pixels wide or tall your panned image was.
In the example shown below, I panned 3 times to the right. The dimensions of the final panned image was 2560x1024 pixels. This is an aspect ratio of 5:2.
I did a 2X zoom out on the panned image and got a square 2X zoomed out image with dimensions 1024x1024 pixels. The reason it is square, is because the aspect ratio of my original starting image was 1:1. When I clicked the 2X zoom out, it used that 1:1 aspect ratio. To keep the 5:2 aspect ratio, I would have needed to click Custom Zoom and set the aspect ratio to 5:2.
You can change the prompt as you pan
If you have Remix Mode turned on in your /settings, you can change your prompt as you pan! With Remix Mode turned on, when you click a pan arrow, you will get a pop-up. In this pop-up, you can change your prompt text.
Below, I've included a set of 4 images to demonstrate this. In the images below, the original square 1:1 image is on the left of the panoramic. The rest of the panorama is different across all 4 images. These differences are related to how I changed the prompt text as I panned to the right.
The prompt for the left side of each image below was /imagine sci-fi cottagecore --v 5.2 --stylize 250 --chaos 35.
Each image caption below contains the prompt text used for every pan to the right I performed.
Panning prompt (same as original): /imagine sci-fi cottagecore --v 5.2 --stylize 250 --chaos 35
Panning prompt: /imagine western desert --v 5.2 --stylize 250 --chaos 35
Panning prompt: /imagine surreal desolation --v 5.2 --stylize 250 --chaos 35
Panning prompt: /imagine jupiter's orbit --v 5.2 --stylize 250 --chaos 35
How many times can you pan in v5.2?
There is a limit to the number of times you can pan.
In the example below, I was able to pan to the right 12 times. At that point, Discord couldn’t show me the next grid results of panned images, but gave me a link to view the results grid. Unfortunately, clicking the link brought me to a http 500 error.
It makes sense that there’s a limit to panning. I've seen at least one other user was able to pan 14 times. I'm not sure yet if panning hits a dimensions or file size limitation. Perhaps this will be upped in the future.
Scroll down to see what prompts I used to get the seasoned pan.
Example of using Midjourney pan while changing the prompt along the way to take us through four seasons
In the image above, I attempted to pan through the seasons. To do this, I changed the text of my prompt at about every 2nd pan. Here was my process:
Make it winter!
I started with the prompt /imagine winter landscape, illustration --v 5.2 --stylize 25 --chaos 20
Now make it spring!
On my second pan, I changed my prompt to /imagine spring landscape, green trees, green grass, illustration ::2 snow::-2 --v5.2 --stylize25 --chaos 20
I really wanted to emphasize that green needed to be a dominant color. I also incorporated weights (the number following the double colon) in my prompt to try to get rid of the snow and emphasize the spring landscape.
Ok, now make it summer!
After a couple more pans, I switched the prompt to /imagine summer pool party with friends, illustration ::2 snow::-2 --v 5.2 --stylize 25 --chaos 20
Now autumn!
After a couple of pans in summer time, I changed the prompt to /imagine autumn landscape, fall colors, red maple trees, autumn forest, illustration ::5 kids::-4 --v 5.2 --stylize 25 --chaos 60 --no kids,people,house,fence
On some initial rolls, MJ really wanted to put another house and more people in the image. To force it more toward a fall landscape, I played with the prompt weights and the --no parameter to downweight specific words further until I got something workable.
Let’s go back to winter!
Lastly, I tried to get it go back to a winter landscape with /imagine winter landscape, snowy winter, blue and white colors, snow, winter ::5 orange::-2 --v 5.2 --stylize 25 --chaos 60 --no orange,red,yellow.
As you can see in the image, I struggled a bit with getting rid of the warm colors, but it did bring back the snow!
At this point, the dimensions of my next results grid were too big for Discord to show me the images and I was unable to pan further.
Summary
Midjourney's latest addition, the 'pan' feature, gives AI art creators the ability to add pixels and expand images, resulting in larger and more detailed final images. Simply click one of the arrow buttons that shows up below your upscaled v5+ image to pan in the direction you want.
With Remix Mode turned on, you can change the prompt text as you pan in any direction. You can use this to mix different styles and themes, creating some really interesting images.
While there does appear to be a limitation to how many times you can pan an image, the ability to customize as you pan, exploring various themes, makes the 'pan' feature a powerful addition to MidJourney.
I hope you enjoyed learning about Midjourney’s pan feature. Thanks for reading!