Resize an image without losing quality: what is possible
Making an image smaller is always safe. Making it bigger never adds detail. How to do both well, the one habit that quietly ruins photos, and the way Voidcanvas keeps a layer's original pixels while you try sizes.
Beginner2 min readUpdated 27 September 2026Editor · Image size, Move tool
What happens to the pixels
| You | What happens | Quality |
|---|---|---|
| Shrink a photo | Pixels are averaged together | Same or better |
| Enlarge a photo up to about 1.5× | New pixels blended from neighbours | Slightly soft; fine for screens and posters seen from a distance |
| Enlarge a photo 2× or more | Most pixels are invented | Visibly soft; edges smear |
| Resize text, shapes, layer styles | Redrawn from their outlines | Sharp at any size |
| Resize the same photo several times | Each resize blends again | Softer every time |
| Save as JPG after each resize | Compression on top of compression | Blocky around edges and text |
In the Editor
- One layer: drag the Move tool's corner handles, or type a new W or H in the Position section of Properties. The layer keeps its original pixels and only stores the new size, so shrinking it and then enlarging it back is lossless, up to the size it arrived at.
- Free transform (Ctrl+T) redraws the pixels when you press Enter. Use it once, for rotation, skew or warp, not for trying sizes.
- The whole design: Image, Image size… (Ctrl+Alt+I) scales every layer. Text and shapes are redrawn; photos are resampled. The Editor saves a version before a large resize so you can go back.
- Exports: 2× and 3× exports redraw text and shapes at the larger size, so a 1080 px design exports a crisp 2160 px file. Photos in it are only as sharp as their own pixels.
When you need it bigger than it is
Ask for the original. A photo from a phone, a camera or a stock site is almost always larger than the copy that was emailed, messaged or screenshotted. If there is no original, use the image smaller than you planned, crop less, or make it a background element under a colour or a texture, where softness reads as intentional. A little Film Grain on top also hides mild softness better than sharpening does.
Does resizing reduce quality?
Shrinking does not, in any way you can see. Enlarging always does, a little at 1.5× and clearly at 2× or more. Repeated resizing and repeated JPG saves reduce it further.
Can AI upscalers enlarge without losing quality?
They produce a sharper-looking result by inventing plausible detail. It is not the real detail, and on faces, text and logos the inventions can be wrong. Use the original when accuracy matters.