Not applicable — a screen size has no physical dimensions
Aspect ratio
16:9
Where this size is used
4K video frames, title cards and thumbnails
High-density displays and digital signage
Desktop wallpapers for 4K monitors and televisions
Masters intended to be downscaled to 1080p later
Resizing to 4K UHD
UHD and DCI 4K are not the same thing
3840 × 2160 is UHD — exactly double 1080p in each dimension, four times the pixel count, and the same 16:9 ratio. Every consumer 4K television, monitor and streaming service means this.
4096 × 2160 is DCI 4K, the digital cinema standard. It is 256 pixels wider and a slightly different ratio, roughly 1.9:1. Cinema cameras and post-production pipelines mean this.
The confusion is real and it costs people renders: deliver DCI to a UHD timeline and it is pillarboxed or cropped. Unless someone has specifically asked for cinema delivery, 3840 × 2160 is the one you want, and it is what this preset produces.
Why doubling is convenient
Because UHD is exactly twice 1080p in each dimension, downscaling a 4K image to 1080p is an exact 2:1 reduction — every output pixel is the average of a clean 2 × 2 block. That is the best-behaved downscale there is, with no resampling artefacts and no fractional pixel positions.
This is a genuine argument for working at 4K even when delivering at 1080p: the master stays useful, and the derivative is free of the softening that a fractional downscale introduces.
Whether your source can fill it
3840 × 2160 is about 8.3 megapixels, which most cameras and phones exceed comfortably — before cropping. Cropping a 4:3 phone photograph to 16:9 discards a quarter of it, so a 12-megapixel source lands at about 9 megapixels, just above the requirement.
Below that, the image is being enlarged. Interpolation makes it bigger without adding detail, and on a 4K screen — which is exactly the kind of display that shows softness — that is visible. A super-resolution model does better on photographic texture and worse on anything containing text.
One warning about the output rather than the input: a lossless 4K image is large. A 3840 × 2160 PNG of a photograph is routinely 15 to 25 MB, where the same frame as a quality-85 JPEG is under 2 MB and visually identical on a screen. Keep the lossless version as a master if you need one, and deliver the lossy one.
Yes. Resizing JPG, PNG, and WebP images is completely free with no sign-up, no watermark, and no file limit. Resizing runs entirely in your browser, so there is nothing for us to charge for. Pro plans add the AI tools, but everyday resizing stays free forever.
Are my images uploaded to a server?
No. OpusImg resizes images on your own device using WebAssembly and your browser's own canvas. Your files never leave your computer or phone — there is no upload, no cloud processing, and nothing stored on our side. That makes it safe for confidential documents, client work, and personal photos.
How do I resize an image to exact pixel dimensions?
Choose the "By dimensions" mode and type the width and height you need in pixels. Leave the aspect-ratio lock on to keep the image's proportions — change one field and the other follows automatically. Turn the lock off when you need an exact, possibly non-proportional canvas, such as a specific banner size.
What is the lock aspect ratio toggle for?
When the lock is on, the width and height stay proportional: setting the width to 1000px updates the height to match the original shape, so the picture is never squashed or stretched. When the lock is off, you can set width and height independently, which is useful for fitting an exact canvas even if it distorts the image slightly.
Which resampling algorithm should I pick?
Use Lanczos for photographs — it produces the sharpest, most natural result and is the default. Bicubic is a fast, smooth alternative handled by your browser. Choose Nearest neighbour only for pixel art, sprites, or icons where you want hard edges and no blending between pixels when scaling up.
What does the DPI setting do?
DPI (dots per inch) is metadata that tells print software how large to print the image. It does not change the pixel dimensions or how the image looks on screen — 72 DPI is standard for the web, while print shops usually want 150 or 300 DPI. We embed the value you choose into the file (PNG and JPG) so a print workflow reads the right density.
Can I resize for Instagram, YouTube, or other platforms?
Yes. The "By preset" mode includes 18 ready-made sizes — Instagram square, portrait, and story; Facebook and LinkedIn covers; X/Twitter headers; YouTube thumbnails and channel art; Pinterest pins; plus HD, QHD, and 4K. Pick a preset and your image is resized to the exact canvas that platform expects.
Will resizing reduce image quality?
Making an image smaller (downscaling) keeps it looking crisp — Lanczos and bicubic both average pixels cleanly. Making it larger (upscaling) can only invent so much detail, so very large enlargements look soft; for big upscales the AI upscaler in Pro reconstructs detail far better. We encode the result at high quality so the resize itself adds no visible compression artifacts.
Can I resize many images at once?
Yes. Drop a whole folder and apply one target to the batch. In percentage mode every image is scaled by the same percentage, so each keeps its own aspect ratio; in dimensions or preset mode every image is fitted to the same canvas. Results are bundled into a single ZIP for download, all processed on-device.
What formats can I resize?
You can drop in JPG, PNG, WebP, AVIF, GIF, BMP, TIFF, and HEIC. Each file is decoded in your browser, resized, and re-saved — JPG and HEIC come out as JPG, PNG/BMP/TIFF as PNG, and WebP/AVIF/GIF as WebP, so the result always opens everywhere. To change format deliberately, use the converter tool linked below.