AI outpainting is a technique where an image model generates new content beyond the original edges of a picture, extending the scene in any direction while matching its style, lighting and perspective. It's also called uncrop, generative expand, image extension or AI zoom-out. The original pixels stay; only the new area is invented.
It's the reason you can now turn a vertical phone photo into a wide banner without cropping anyone out. This explainer covers how it works, what it's good and bad at, and how it differs from related techniques. For practical steps, see how to change an image's aspect ratio without cropping.
How outpainting works
Modern image models (diffusion models and newer multimodal models such as Google's Gemini image models) learn what real images look like from huge training sets. Outpainting gives the model:
- A larger canvas with your image placed inside it.
- A mask marking which areas are empty and need filling.
- Optionally, a text prompt describing what should be there ("more beach and sky").
The model then generates content for the masked area that is consistent with the visible part. It continues lines, repeats textures, extends shadows in the same direction and keeps the colour grading. Good tools then blend the seam so you can't see where the original ends.
At AImageResize we use Google's Gemini image models for this. In practice, the quality depends less on the "magic" and more on what's at the edge of your photo.
Outpainting vs inpainting vs upscaling
These three get confused constantly:
| Technique | What it changes | Example |
|---|---|---|
| Outpainting | Adds new area outside the image | Portrait photo to 16:9 banner |
| Inpainting | Replaces an area inside the image | Removing a person from the background |
| Upscaling | Adds pixels to the same picture | 1000 px photo to 4000 px |
| Cropping | Removes area | 16:9 photo to square |
They combine well. A typical workflow is outpaint to the right shape, inpaint any odd detail the AI invented, then upscale if the final size needs more resolution. Our guide to resizing without losing quality covers the upscaling part.
What outpainting is good at
- Changing aspect ratio for different platforms: one photo to Instagram 4:5, Story 9:16 and YouTube 16:9.
- Fixing tight framing, like a product too close to the edge or a head cut off at the top.
- Making room for text on ads, banners and thumbnails.
- Turning photos into wallpapers for desktop (16:9) or phone (roughly 9:19.5).
- Converting portrait to landscape, the single most common request. See how to convert a portrait photo to landscape.
Where outpainting struggles
Be realistic about these:
- Text and logos at the edge. Models extend letters into gibberish.
- People cut off at the frame. The AI has to invent the rest of the body, and hands are still hard.
- Strict repeating patterns like tiles, fences and brick walls can drift out of alignment.
- Very large extensions. Doubling or tripling the canvas gives the model less to go on. Extend in steps.
- Factual content. Outpainting invents plausible scenery, not the real one. Don't use it on news photos, evidence, or a marketplace product shot where the area shows part of the product.
Is outpainted content "real"?
No, and you should treat it that way. The new area is a plausible guess. That's fine for a banner background, and not fine if it implies something untrue. On marketplaces, keep the product itself exactly as photographed. Our guides on Amazon image requirements and Etsy listing photos explain why.
Which tools offer outpainting?
Outpainting is now built into many editors: Photoshop (Generative Expand), Canva (Magic Expand, on paid plans), plus dedicated web tools. We compare them in best AI image extenders.
AImageResize focuses on one job: type the size you need and AI fills the gap. Sign in with Google for 3 free AI generations, or read the complete guide to resizing images without cropping.
Frequently asked questions
Is outpainting the same as "uncrop"?
Yes. Uncrop, generative expand, AI image extender and zoom-out are all names for outpainting.
Does outpainting reduce image quality?
The original pixels aren't changed. The new area is generated at the output resolution, so it usually matches the original's sharpness. Look closely at the seam to check.
Can I control what the AI adds?
Many tools accept a text prompt describing the new area. Even without a prompt, the model continues what it sees at the edges, so a beach photo gets more beach.
Is AI outpainting free?
Most tools offer a few free generations, then charge credits or a subscription, since every generation costs the provider GPU time.