How to Use Midjourney's Pan and Zoom Out to Extend Compositions
Midjourney's pan and zoom out tools solve a specific problem: you have a strong image at one size or crop, and you want to see more of the scene without generating from scratch. These tools extend the image outward, generating new content that continues the original. For wallpaper creation specifically, they are useful for converting a portrait-oriented output into a landscape composition, or for creating wider images than Midjourney's standard aspect ratios produce in a single generation.
Understanding what these tools actually do, and where they produce coherent vs incoherent results, is what separates useful extensions from wasted generations.
Zoom Out: Pulling Back From the Original Image
Zoom Out places the original image in the center of a wider canvas and generates new content around it. You access it by upscaling an image and then clicking Zoom Out 2x, Zoom Out 1.5x, or Custom Zoom from the button row below the upscaled result.
What happens during a zoom out
The original image is placed at the center of the new canvas, scaled down to a proportion of the new image size. Midjourney generates the surrounding area based on context from the original image and your current prompt. The zoom out level determines how much new content is generated relative to the kept original.
- Zoom Out 1.5x: The original occupies approximately 44% of the new canvas area. A moderate amount of new content is generated around it. Less extension, more coherent results.
- Zoom Out 2x: The original occupies approximately 25% of the new canvas area. Significantly more new content is generated. Results are more variable and sometimes introduce elements that conflict with the original scene.
- Custom Zoom: Lets you set a custom zoom value and change the aspect ratio of the extended image. This is particularly useful for extending a 1:1 square image to a 16:9 landscape, or a 9:16 portrait to a 1:1 square.
What works well with zoom out
- Scenes with clear spatial logic (landscape with horizon, architectural interior, open space with consistent depth) that the model can extend plausibly
- Images where the original content is centered and the extended areas are background or continuation of existing elements
- Abstract or atmospheric images where continuity does not require precise object placement
What produces poor zoom out results
- Images with specific foreground objects near the edges that would need to continue naturally into the extension but cannot (a figure whose body extends to the edge, a building whose architecture continues off-frame)
- Images with very distinctive lighting or atmospheric conditions that are hard for the model to match in the extension area
- Extremely detailed textures at the original image boundary that the extension cannot seamlessly continue
Pan: Extending in a Single Direction
Pan adds new content on one side of the image: left, right, up, or down. The original image is kept and new content is generated on the selected side. You access pan with the directional arrow buttons (left, right, up, down arrows) that appear below an upscaled image.
How pan differs from zoom out
Pan extends in one direction only, keeping the original image at its full size and appending new content to one edge. This is more useful than zoom out for creating wider panoramic compositions because the original image retains its full resolution and detail rather than being scaled down. You can pan repeatedly to create progressively wider images.
Aspect ratio changes with pan
Each pan operation adds a strip to the chosen side, changing the overall aspect ratio of the combined image. A 1:1 square panned left becomes approximately 4:3. Panned again becomes approximately 16:9. This is a useful progression for building landscape wallpapers from portrait or square starting points.
What works well with pan
- Landscape scenes where the world extends naturally in the horizontal direction: mountains, cityscapes, forests, open terrain
- Scenes with a clear horizon line that gives the model a spatial anchor for what to generate on the extended side
- Abstract atmospheric content where the extension needs to continue a mood rather than specific objects
What produces poor pan results
- Images where the edge you are extending from contains a specific object or figure that needs to continue naturally (a face at the left edge being panned left will produce an awkward continuation)
- Very tight close-up compositions where the model has insufficient spatial context to understand what should continue off-frame
- Images with strong edge lighting (a hard light beam hitting the edge of the frame) that cannot be naturally continued in the extension
Modifying the Prompt During Extension
Both pan and zoom out allow you to modify the prompt for the extension operation. This is an underused feature that can significantly improve results.
When extending a landscape to the right and you want to introduce a specific element in the extension area (a city in the distance, a body of water, a mountain range), add that element to the prompt before executing the pan. The model uses both the visual context of the original edge and your text prompt to generate the extension. Without prompt modification, the model generates a continuation based purely on visual context inference, which is often good but less controlled than guiding it with text.
Example: if the original image is a forest at dusk, and you pan right, adding "clearing with distant mountains, fading light" to the prompt before panning gives the model direction for what the continuation should show rather than just more forest.
