Image Resizer

Image resize

Upload requirements are never consistent — forum profile pictures cap at 500×500, LinkedIn banners expect 1584×396, print jobs want 300 dpi at final size. This resizer takes any image and scales it to exact pixel dimensions or a percentage, with a choice of resampling algorithms (bicubic, Lanczos) so sharp details survive the shrink.

How to resize an image

  1. 1

    Upload the image

    Any JPG, PNG, WebP or HEIC.

  2. 2

    Pick dimensions

    Exact pixels, percentage, or "fit within" a box while preserving ratio.

  3. 3

    Pick resampling method

    Lanczos for photos (default), nearest-neighbour for pixel art, bilinear for speed.

  4. 4

    Export

    Output in the same format or convert on the way out.

Resampling algorithms compared

Different resize algorithms sample surrounding pixels differently.

Algorithm guide

Method How it samples Best for
Nearest neighbour Copies the closest pixel Pixel art, retro graphics
Bilinear Weighted average of 4 pixels Speed; acceptable for modest resizes
Bicubic Weighted average of 16 pixels General-purpose photo resize
Lanczos (a=3) Uses sinc kernel over a ring of pixels High-quality photo downscaling
Mitchell Compromise between smooth and sharp Balanced default

Common resize targets

Target Dimensions
LinkedIn profile 400 × 400
LinkedIn banner 1584 × 396
Instagram square post 1080 × 1080
Instagram story 1080 × 1920
Twitter profile 400 × 400
Facebook cover 820 × 312
YouTube thumbnail 1280 × 720
Zoom virtual background 1920 × 1080
Discord server icon 512 × 512

Upscaling caveats

Enlarging a small image beyond its original size always loses detail — no algorithm invents information that was not there. Lanczos produces the least artefacting, but for dramatic upscales (2x and beyond), use an AI upscaler that actually synthesises plausible detail.

Aspect ratio

“Fit within” mode preserves the ratio and may leave one dimension smaller than the target. “Cover” mode crops to fill exactly. Pick based on whether keeping the full image matters more than matching dimensions exactly.

Privacy

All resampling runs in your browser. No upload, no server-side processing.

Frequently Asked Questions

Lanczos for downscaling photographs, bicubic for modest upscales (up to about 1.5x), nearest neighbour for pixel art and game sprites. For heavy upscaling, no classical method will look good — use an AI upscaler instead.

Yes — drop multiple images and the same target dimensions apply to each. Useful for preparing a whole photo set for a particular platform.

Roughly. Halving dimensions (50% each way) quarters the pixel count and typically drops file size by 70-85% for JPGs, depending on how much fine detail the smaller version retains.

The resampler operates in the image’s colour space; sRGB in and out. Gamma-correct resampling (which treats pixel values linearly during the interpolation) is an advanced toggle — on by default for accurate highlight preservation, off for speed.