Leave the subject whole
The AI continues what it can see. A subject shown whole, with some space around it, extends most convincingly, so pick a photo where the person or product sits fully inside the frame.
Pick a new shape and the AI image extender continues the scene beyond the frame. Your original photo is laid back over its own area, so it keeps its detail, and a run gives you 2 results by default (1 to 4) at 2K or larger.
Try one of these
Real results from this tool. The dashed frame marks the original photo; everything outside it was generated.



A square lake photo extended to a 16:9 desktop picture, a 9:16 story and a 3:1 banner. The cabin and the lake inside the frame are the original photo.
Original
A portrait on a stone bridge turned into a 16:9 cover. One run gave these two different extensions; keep the one that fits your design.




Choose a shape and the image extender fills in the rest; add a short direction when you want something specific.
Drop, paste or choose a JPG, PNG or WebP up to 10 MB.
Pick 1:1, 4:5, 16:9, 9:16, 3:2, 2:3 or a 3:1 banner, and how many results you want. Add a short direction if you want something specific in the new area.
Results come back in about half a minute. Keep the one you like as a PNG, or as a layered PSD with your original on its own layer.
Six strengths, built around one promise: the image extender adds the new area and keeps your photo.

OriginalAfter generation your original is laid back over its own area and blended only along a narrow seam, so the new parts are AI and your photo keeps every detail. Landscapes, skies, water, walls and floors simply continue.

A run gives 2 different extensions by default, or pick 1 to 4, so you choose the one that fits your design. Each result costs 2 credits, shown on the button before you start.

OriginalResults are at least 2048 px on the long side for most photos and up to 5000 px, ready for slides, banners and print. Small photos are enlarged so the result reaches about 2K.

1:1, 4:5, 16:9, 9:16, 3:2, 2:3 and a 3:1 banner turn one photo into post, story, cover and banner sizes. The new size is shown before you run.

Download a PNG, or a PSD with the generated picture and your original on its own masked layer, ready to fine-tune in Photoshop.

OriginalPictures stay in your own history behind unguessable links, visible only to you, and stay out of AI training.
Five short tips for extended pictures that look like they were shot that way.
The AI continues what it can see. A subject shown whole, with some space around it, extends most convincingly, so pick a photo where the person or product sits fully inside the frame.
Use 9:16 for stories and Reels, 4:5 for feed posts, 16:9 for YouTube and slides, and 3:1 for website and Twitter/X headers. The tool centers your photo and fills both sides evenly.

The direction box is optional. A few words such as "more beach and sky" or "plain wall for text" steer the new area, and you can type in any language.
Each result is a different take. Two results usually give one that fits, and another run brings new variations. Your original is laid back on top every time, however many times you extend an image.

The layered PSD keeps your original on its own layer, with a mask over the generated picture. Paint the mask to show more or less of either and blend the seam exactly as you like.
Wherever one photo has to fit many sizes, an image extender keeps the whole subject in every one.

One photo as a feed post, a story and a cover without cropping the subject.

Product shots with room for prices, badges and banner text.

Wide 16:9 backgrounds for presentations and YouTube thumbnails.

Hero banners, headers and posters from photos that were cropped too tight.
Answers about price, shapes, sizes, your original, prompts and privacy.
It places your photo on a larger canvas in the shape you choose and generates the missing area so it continues the scene: light, perspective and texture. Here your original is then laid back over its area, so only the new part is generated.
2 credits per result. A run gives 2 results by default (4 credits), and you can pick 1 to 4 (2 to 8 credits).
It uses credits: each result runs on a paid AI service, 2 credits a result, and you sign in first. The price is shown on the button before you start, and if a run fails the credits are returned.
Your original is laid back over its own area after generation and blended only along a narrow seam (12 to 64 px), so faces and objects inside your photo keep their detail. A subject with some room around it extends most naturally, and small photos are enlarged to reach 2K.
Yes. Choose 16:9 or 3:2 and the image extender centers your photo and generates both sides. A photo with open sky, water or a plain wall at its sides gives the most natural landscape.
1:1, 4:5, 16:9, 9:16, 3:2, 2:3 and a 3:1 banner: 4:5 for feed posts, 9:16 for stories and Reels, 16:9 for YouTube thumbnails and slides, 3:1 for headers. Results are at least 2048 px on the long side for most photos and up to 5000 px; the exact size shows before you run.
Yes, if you want to. Type a short direction such as “more sky” or “plain wall for text”; it guides only the new area. You can write it in any language.
They go only to the AI services that check and extend them, and stay in your own history behind unguessable links, private to you and out of AI training. You can delete your runs from My creations at any time.
Upload a photo, pick a shape and choose from 2 results, with your original laid back on top.
Extend an image