How to resize an image
Open the Image Converter, load the image, and set either a percentage or exact pixel dimensions. Leave the aspect ratio locked and enter one dimension; the other follows automatically. Convert and download.
Resizing resamples the image, which changes pixel values even when the output format is lossless. Making an image smaller discards detail permanently, and making it larger cannot add detail that was never captured — an enlarged photo is an interpolation, not a recovery.
Last reviewed: 2026-09-19. Checked against the tool’s own source on that date.
Percentage or pixels?
Use a percentage when you are scaling a batch of differently sized images by the same amount, or when you just want "about half". Percentage scales both axes by the same factor, so the aspect ratio is preserved by construction.
Use exact pixels when something downstream has a specific requirement: a column width, an upload limit, a print dimension. Enter the dimension you care about and let the other follow.
If you unlock the aspect ratio and set both dimensions, the picture is stretched. That is occasionally deliberate and almost always a mistake, so it is a separate, explicit action rather than something that can happen by accident.
What resizing actually costs
Downscaling maps many source pixels onto one output pixel. The detail in the discarded pixels is gone, and scaling back up afterwards will not bring it back — you will get a blurred approximation of what was there.
Upscaling invents nothing. Every enlargement algorithm interpolates between the pixels it already has, so a 400-pixel image stretched to 1600 pixels has the same amount of real information spread over sixteen times the area. It will look soft, and no setting changes that.
This means one practical rule: keep your original, and resize from it every time. Repeatedly resizing an already-resized file compounds the softening, and if the format is lossy it compounds compression artefacts as well.
The order of operations, and why it matters
The Image Converter applies crop first, then resize, then rotation. That order is fixed and it is the one that behaves predictably: cropping first means the resize percentage refers to the part of the image you kept, not to the original frame.
If you crop 50% of the width away and then resize to 50%, you get a quarter of the original width — which is what most people expect. Resizing first and cropping second would give a different result from the same two numbers.
Choosing a target size
- For a web page, aim for roughly twice the CSS width the image will occupy, to look sharp on high-density screens. Beyond about twice, the returns effectively stop.
- For print, the dimension that matters is pixels divided by the printed size. Around 150 pixels per inch is fine for most documents; 300 is the traditional target for photographic printing.
- For an email attachment or a form with an upload cap, resize first and only then adjust quality — dimensions are the stronger lever.
Worked example: a 3000-pixel screenshot for a 600-pixel column
A screenshot is 3000 by 1875 pixels. It will be displayed in a 600-pixel-wide column. Storing it at 3000 pixels wastes bandwidth and, on some devices, memory.
Doubling the display width for high-density screens gives a target of 1200 pixels wide.
- Load the screenshot into the Image Converter.
- Choose resize by pixels and enter 1200 for the width, aspect ratio locked.
- Confirm the height fills in as 750 automatically.
- Keep PNG as the output, because a screenshot has sharp text and flat areas.
- Convert and download.
Result: A 1200 by 750 PNG. The pixel count falls from 5.6 million to 0.9 million, about 16% of the original, and the text stays crisp because PNG does not soften edges the way a lossy format does. The original 3000-pixel file is untouched on your device and is what you should come back to next time.
Open the tool
Resize an image in the Image Converter
Resizes by percentage or exact pixels in your browser, with the aspect ratio locked by default. It does not upscale intelligently.
What this does not cover
- Enlarging cannot add detail. The tool will do it, and the result will be soft; that is a property of the mathematics, not of this implementation.
- Any resize resamples the image, so even PNG output is no longer bit-identical to the source.
- Very large images are clamped to a pixel budget your device can safely allocate, so an extremely large requested output may come back smaller. The tool says when this happens.
- Input is JPEG, PNG, WebP or GIF only.
- There is no content-aware scaling or smart cropping. Resizing is uniform across the whole frame.