Image tool
Resize image
Make a picture the size something asked for. It is scaled here, in this page, using a stepwise method that keeps detail rather than the single-pass one that makes downscaled photos look like a fax.
Drop an image here
Or press to choose one.
nothing is uploaded — this page does the work
Why downscaled photos usually look bad, and why these do not
Drawing a 4,000-pixel-wide photograph into a 400-pixel-wide box in one step makes the browser pick one source pixel out of every ten and throw the other nine away. Fine detail turns into noise and straight edges turn into staircases. It is the single reason quickly-resized images look cheap.
This tool halves the image repeatedly until it is within a factor of two of the size you asked for, then does the final step. Each halving averages four pixels into one, so nothing is discarded without being taken into account. It costs a few milliseconds and it is the difference between a photograph and a smear.
Width, height, or percentage
Type into any one of the three and the others follow, as long as keep the proportions is ticked. Untick it to squash or stretch — occasionally what you want, usually not.
Never enlarge is on by default and should usually stay on. Asking for 4,000 pixels from a 900-pixel image does not create detail that was never captured; it creates a soft, larger version of the same picture and a much bigger file. If something has asked you for a 4,000-pixel image and you only have 900, the honest answer is that you need a better source.
Which format to save as
JPEG for photographs. Small, universally supported, no transparency.
PNG for screenshots, diagrams, logos and anything with flat colour or transparency. Lossless, and much larger for a photograph.
WebP when you control where the image will be shown. Roughly 25-35% smaller than JPEG at the same visual quality, and it handles transparency. Every current browser reads it; some older desktop software still does not.
The quality slider affects JPEG and WebP. PNG ignores it, because PNG has no lossy setting to turn.
What this strips out, deliberately and otherwise
Putting an image on a canvas and reading it back gives you the pixels and nothing else. Every piece of metadata is gone from the result: the EXIF block, and with it the GPS coordinates where the photo was taken, the camera's serial number, the date, the lens, and any copyright field.
For most people that is a privacy improvement they did not know they needed — a photo posted straight off a phone frequently carries the exact location of the house it was taken in. If you are publishing pictures, this is a feature.
If you are a photographer, it is a real loss: the embedded colour profile goes too, so a wide-gamut file comes out interpreted as sRGB and may look flatter. For colour-managed work, resize in software that understands profiles. This tool is honest about the limitation rather than quiet about it.