Is the original modified?

No. The prompt requires the original region to keep its own scale with no crop, zoom or re-render. Only the newly added surroundings are generated.

Is the generated content accurate?

It is a plausible continuation inferred from your image, matched in style and lighting — not a record of what was really there. Fine for creative and layout work; be careful with documentary use.

Can I say what to add?

Yes — use the optional hint field, e.g. "more meadow and distant hills on both sides".

Which ratios are supported?

Six: 16:9, 9:16, 1:1, 4:3, 3:4 and 21:9, covering the sizes the major platforms ask for.

How far can it expand?

That depends on the gap between the original ratio and the target. The larger the gap, the more the model has to invent, so avoid very large jumps in one pass.

What resolution comes out?

Always 2K, which is plenty for web, social and ordinary print.

AI Image Expander

Upload an image and choose a target ratio. The AI continues the scene outwards — painting in what should exist beyond the frame instead of cropping or stretching what you already have.

Original never croppedSix target ratiosInvisible seamsFits every platform

Upload an image

0/1

Click to upload or drop a filepng, jpg, jpeg, webp (1 left)

🎯Choose a target ratio

📝What should fill the new space (optional)

0/5000

Required: 20Available: 0

Ready to expand

Upload an image and choose a target ratio

Expansion results

Samples from GPT Image 2 — the original area is untouched, the surroundings are generated.

Vertical to widescreen

Square to 16:9

Portrait with room added

Landscape to panorama

Product with margin added

Out to cinematic 21:9

What the expander does well

The point is not filling space — it is that nobody can tell which part was filled.

Not one original pixel moves

The prompt explicitly keeps the original at its own scale — no crop, no zoom, no re-rendering of the subject. Only the previously unseen surroundings are new.

Seams you cannot find

Generated areas continue the existing perspective, horizon, lighting direction, colour grade, grain and texture, so the whole frame reads as a single photograph.