How to extend a photo's background: uncrop to any shape

What extending a photo actually does

When you extend a photo, the original pixels are kept and placed inside a larger canvas. The empty area around them is filled with new content that continues the scene: more sky above a landscape, more wall beside a portrait, more sand and grass along a beach path. The result is a new picture in the chosen shape, with your original photo at its centre.

The new area is invented, not recovered. The AI does not know what was really beyond the frame, so it paints something plausible. That is perfect for backgrounds like sky, sea, walls, floors, fields and foliage, and much less reliable for people, text or detailed objects that were cut off at the edge.

Choose a photo that extends well

• Leave space around the subject. A person or product fully inside the frame extends cleanly; a body cut off at the edge forces the AI to invent the missing half.
• Simple, continuous backgrounds work best: sky, sea, plain walls, grass, sand, a blurred street.
• Busy scenes with crowds, signs, patterns or architecture right up to the edge are much harder to continue convincingly.
• Use a sharp photo. Blur and heavy noise get continued into the new area; fix a soft shot first with Photo deblur.
• Remove distractions near the edges before extending, for example with the object remover, so they are not repeated.

Step by step: extend a photo with AI

1. Upload your photo

   Open the Image expander and upload a JPG, PNG or WebP of at least 128 px a side. Larger photos are scaled to at most 1280 px before upload.

2. Pick the new shape

   Choose 1:1 Square, 4:5 Portrait post, 16:9 Widescreen or 9:16 Story. After signing in you see the exact credit price before the run.

3. Run the expansion

   It usually takes 10 to 40 seconds. Failed generations aren't charged — held credits come back automatically. Your result is kept for 24 hours to download.

4. Check the edges

   Look closely at where the original photo ends and the new area begins, and at anything that was cut off: hands, feet, hair, furniture or lines such as a horizon.

5. Run again or adjust the source

   If the new area looks wrong, run it again, try a less extreme shape, or crop the source photo so the subject is no longer cut off at the edge and extend that.

Which shape for which use

• 16:9 Widescreen: website headers, presentation slides, video thumbnails and cover images. Turns a portrait photo into a wide banner with room for a headline.
• 9:16 Story: stories, reels covers and phone wallpapers. Turns a landscape shot into a tall picture.
• 4:5 Portrait post: feed posts that take up more screen height than a square.
• 1:1 Square: profile pictures, marketplace listings and square feed posts, without cropping a tall or wide photo.
• Extending to the opposite orientation, such as tall to wide, adds the most new area. The further you go, the more the result depends on how simple the background is.

Limitations and what to check

The extended area is AI-generated and varies from run to run. Repeated patterns, a horizon that does not line up, doubled objects or odd shapes near the original edge are the most common problems, so always check the full picture at 100% before you use it.

• Your whole photo keeps its size and new area is added around it, up to about 2300 px on the long side. For a larger file, run the result through the AI upscaler.
• People or text near the original edge may come out distorted; keep them inside the frame where you can.
• Do not use extended photos where an unedited original is required, such as documents, evidence or news.
• Only upload photos you have the rights to, and photos of other people only with their consent.

Frequently asked questions