Skip to content
Reembun
Image compressor icon
Image

Image compressor

Compress images by quality or to a target file size, in batches, with the original never leaving your device.

Drag files here, or

JPEG, PNG, WebP, AVIF and GIF. Decoded on your device, never uploaded.

Encodes at quality 76, tuned to be indistinguishable from the original at normal viewing size. This is the setting to use when you just want the file smaller.

px

Optional. Height follows automatically.

Ready. Runs locally on your device.

Your files stay on your device. The tool works directly in your browser, using your device to process your files. Nothing is sent to our servers, and we never receive, store, or see your files or figures.

Share

Share this page

Share result

Your results

Use the tool above first. Whatever it works out shows up here, ready to copy or share.

Cite

Cite this page

Reembun. (2026, July 31). Image compressor. https://reembun.com/image-compressor
Pick a style, then copy the reference. The access date is today.

How to use it

  1. Add your images

    Drop them on the box or use Choose images. Several at once is fine, and none of them is uploaded.

  2. Choose how hard to compress

    Smart picks a quality for you, By quality hands you the slider, and To a file size works backwards from a target such as 200 KB.

  3. Set the output

    JPEG for photographs, WebP or AVIF for the smallest files, PNG when you need transparency. Limit width shrinks oversized photos in the same pass.

  4. Check, then download

    Each image shows its original and compressed size side by side, so you can back off the compression before you save anything.

How compression works here

Your image is decoded to pixels in your browser and re-encoded at your chosen quality, all inside the tab. Nothing is transmitted at any stage.

JPEG and PNG are handled by dedicated encoders running inside the page, the same class the well known upload based compressors run on their servers. They produce noticeably smaller files than the encoder built into your browser: roughly 20% smaller for JPEG at matched quality, and far more than that for PNG. WebP and AVIF use the browser’s own encoders, which are already the best available for those two formats.

For lossy formats such as JPEG, WebP and AVIF, the quality setting controls how aggressively fine detail is discarded. The encoder works on 8×8 blocks of the image, keeping the coarse structure your eye notices and throwing away the fine variation it does not. Lower quality means more discarded, smaller files, and eventually visible blocking around edges.

PNG works differently. It has no notion of discarding detail, so the only real lever is how many distinct colours the file is allowed to contain. Dropping from millions of colours to a few hundred is invisible on a screenshot or a logo and cuts the file dramatically. On a photograph it is the wrong tool, and JPEG or WebP will serve you better.

Choosing a format

FormatBest forTypical saving vs JPEG
JPEGPhotographs, anything without transparencybaseline
WebPAlmost everything; supports transparency25-35% smaller
AVIFMaximum compression, modern browsers40-50% smaller
PNGScreenshots, logos, flat colour, transparencystill larger for photos

WebP is the pragmatic default in 2026, supported by every current browser, meaningfully smaller than JPEG, and it handles transparency so it can replace PNG for many graphics too.

AVIF compresses best but encodes slowly, particularly on phones. For a batch of fifty images that difference is felt.

Choosing a compression mode

Smart is the one to reach for by default. It picks the setting per format, and for PNG it reads the image first and chooses the smallest palette that keeps it looking untouched. This is the equivalent of what the popular upload based compressors do when you drop a file on them, and it is right the large majority of the time.

By quality is right when you want consistent visual fidelity across a set of images and do not care what each one weighs.

To a file size is right when something external imposes a hard limit: a government portal capping uploads at 1 MB, a job board at 500 KB, an email attachment budget. The tool encodes repeatedly, bisecting the quality setting, until it finds the highest quality that fits under your ceiling. Eight iterations land within a fraction of a percent of the ideal.

Resizing is the bigger lever

Compression quality is a modest dial compared with dimensions. A 4,000-pixel-wide photo displayed in a 800-pixel column carries 25 times more pixels than it needs.

ChangeTypical size effect
Quality 95 → 80~50% smaller
Quality 80 → 60~30% smaller
Width 4000px → 1600px~84% smaller

If a file is far too large, cap the width first and adjust quality second. The width limit in this tool applies before encoding, so the two work together.

When an image gets larger

Re-encoding an already-compressed image at a higher quality than it was originally saved with adds bytes without adding detail, because the lost information is gone and cannot return. If the tool reports a size increase, the original was already compressed harder than your current setting. Lower the quality, or keep the original.

For PNG specifically, a quality of 100 means lossless, and a lossless re-encode of an already optimised PNG can come out slightly larger. Move the slider below 100 and the palette reduction will almost always win the bytes back.

Generation loss is real: every save-and-recompress cycle degrades a lossy image a little further. Always compress from the highest-quality original you have, never from a file that has already been through the process.

Metadata

Re-encoding drops EXIF data: camera model, lens, settings, and any GPS coordinates recorded at capture. For publishing to the web that is a privacy benefit, since location data in a photo is a genuine disclosure. If you need the metadata preserved, keep your original file alongside the compressed one.

Common questions

How is this different from other compressors?

Your photos are never uploaded. Every other popular compressor sends your image to a server, processes it there and sends it back, which means a copy existed on someone else's machine, however briefly. This one decodes and re-encodes entirely in your browser, so you can compress client photos, ID documents or anything else confidential without a transfer ever happening.

What does the Smart preset do?

It chooses the settings for you, per format and per image. For JPEG, WebP and AVIF it encodes at the quality that is indistinguishable from the original at normal viewing size. For PNG it counts the colours in your image first. If there are 256 or fewer, which is typical of screenshots and logos, it uses exactly that many, so the result is mathematically identical to the original and still much smaller. Only richer images have their palette reduced. If you would rather set the dial yourself, switch to "By quality".

Does the quality slider do anything for PNG?

Yes. At 100 the PNG stays lossless, exactly as before. Below 100 the image is reduced to a colour palette, which is the only thing that shrinks a PNG substantially and is the same technique the well known online compressors use. Screenshots, logos and flat colour graphics usually survive this with no visible change at all, often at a third of the original size. Photographs and smooth gradients are the risky case, since a reduced palette can show banding.

What quality should I use?

80 is the standard answer for photographs and is genuinely hard to distinguish from the original. Below 60, artefacts become visible in smooth gradients like skies. Above 90, file size grows steeply for almost no perceptible gain. Detailed textures tolerate lower quality than flat areas do.

Can I compress to an exact file size?

Yes, for JPEG, WebP and AVIF. Switch to "To a file size" and enter a target in KB. The tool bisects the quality setting until the result fits, which is what you want when an upload form rejects anything over 2 MB or a job application caps a photo at 500 KB. PNG is not offered here because searching for the right palette size would mean running the quantiser repeatedly, which is too slow on a large image.

Last reviewed