Compress & convert images
right in your browser

What happens to an image, step by step

Your file is decoded into a bitmap, drawn onto a canvas at whatever dimensions you asked for, and re-encoded by the browser's own image encoder. That is the whole pipeline, and its consequences are worth spelling out because they explain almost every result you will see.

The decode step reads EXIF orientation and applies it, so a photo taken with the phone rotated arrives upright rather than on its side. The draw step uses high-quality smoothing, which matters when you scale down — it averages source pixels rather than dropping them, which is the difference between a clean reduction and a jagged one. The encode step is where the size comes from: JPEG and WebP take your quality setting, PNG ignores it entirely.

Because the image is rebuilt from raw pixels, everything that was wrapped around those pixels is gone from the output — EXIF tags, GPS coordinates, capture time, camera and lens data, and any embedded colour profile. For sharing that is a privacy win, and it is why the rotation has to be baked in rather than left as a tag. It also means that if a photo was stored in a wide-gamut profile such as Display P3 or Adobe RGB, its colours may shift once that profile is discarded — noticeable on saturated reds and greens, invisible on most everyday photographs.

Which format to choose

Target size works differently from the quality slider. Rather than fixing quality and accepting whatever size results, it searches for the highest quality that still fits under your limit — narrowing the range across several encoding passes. If even a low quality setting cannot get under the target, it starts scaling the image down in 20% steps instead. It is the setting to use when a form imposes a hard cap. It does not apply to PNG, which has no quality dial to search.

Limits and honest trade-offs

Nothing here is lossless for JPEG. Even at 100% quality the file is decoded and re-encoded, so a small amount of information is discarded. That makes compression generational: compressing an already-compressed photo compounds the damage, and each round is irreversible. Always work from the largest original you have rather than re-squeezing a file this tool already produced.

The quality slider does nothing for PNG. PNG compression is lossless, so there is no quality to trade. Running a photographic PNG through as PNG can even produce a larger file than the original, which is why the badge sometimes shows a plus rather than a minus — the tool reports growth honestly instead of quietly keeping the original. If a PNG needs to be smaller, either resize it or convert it to JPG or WebP.

Memory is the real ceiling. A decoded bitmap costs four bytes per pixel regardless of how small the compressed file was, so a 12-megapixel photo occupies about 49 MB while it is being processed and a 48-megapixel one about 195 MB. Results are also held in memory until you clear them. Large batches of high-resolution photos are where a phone runs out of room.

This is the browser's encoder, not a specialist one. Dedicated tools built on mozjpeg, oxipng or similar squeeze out extra bytes at the same visual quality by doing far more work per file. The trade here is speed and privacy: nothing is uploaded and nothing waits in a queue, and for the vast majority of images the browser's encoder is entirely adequate. Animated GIFs are a genuine gap — only the first frame survives, since the canvas holds one image.

Getting the biggest savings

Resize before you reach for quality. This is the advice that changes results most and the one people skip. File size scales with pixel count, so halving both dimensions removes three-quarters of the pixels — a far larger saving than any quality adjustment, and with no visible loss if the image is still larger than it will ever be displayed. A 4000-pixel-wide photo going into a blog post that renders it at 800 pixels is carrying twenty-four times more pixels than it needs.

Then set quality by destination. 80% is the default here because it is where most people stop being able to see a difference. Drop to 70% for thumbnails and previews; go to 90% for images that will be viewed large or edited further. Below about 60% you will see blocking in skies, gradients and shadows — the compare slider is there to make that call for yourself, and it is worth using at least once at full size rather than trusting a number.

Match the format to the content. Photographs belong in JPG or WebP; screenshots, diagrams and anything with text or flat colour belong in PNG or WebP, where JPEG's ringing artefacts around sharp edges are obvious and ugly.

Since the compressed file has no metadata at all, this is also a practical way to strip location data before posting a photo — though if that is your actual goal, the EXIF remover shows you what was in the file before removing it, which is more reassuring than assuming. For single-purpose runs there are dedicated pages for JPG, PNG and WebP, and the full set is on the image tools page.