How to reduce image file size without wrecking it
Updated 2026-08-26 Β· about 7 minute read
A photo straight off a modern phone is often 4β12 MB. Emailed, uploaded or dropped into a document, that is far more than anyone needs β and it is usually the thing making your page slow or your attachment bounce. The good news is that almost all of that size is removable without any visible difference, provided you pull the right lever.
Why your photos are so large
Two reasons compound. First, phone cameras capture enormous images β 12 megapixels is around 4000Γ3000 pixels. Second, they compress lightly to preserve editing headroom.
The mismatch is that almost nothing displays an image at 4000 pixels wide. A full-width photo on a website is typically 1200β2000 pixels. An email attachment someone views on a laptop needs maybe 1600. Everything beyond that is downloaded, decoded and thrown away.
Lever one: dimensions
This is the one people forget, and it is by far the most powerful, because file size scales with the area of the image rather than its width.
Halving the width and height quarters the pixel count. Going from 4000Γ3000 to 2000Γ1500 removes 75% of the data before compression does anything at all. Going to 1000Γ750 removes 94%.
So the first question is never "how much can I compress this?" β it is "how large does this actually need to be?" Resize first, compress second. Doing it the other way round leaves most of the saving on the table.
One caution: scaling up cannot add detail that was never captured. Enlarging a small image just produces a soft, larger one. Always start from the biggest original you have.
Lever two: compression quality
JPEG and WebP are lossy: they discard information the eye is poor at noticing. The quality setting controls how aggressively.
- 90β100 β near-identical to the original, but files stay large. Rarely worth it.
- 80β85 β the sweet spot. Visually indistinguishable from the original for almost all photographs, at a fraction of the size.
- 60β75 β noticeably smaller; artefacts start appearing in smooth gradients like skies.
- Below 60 β visible blocking and colour smearing. Only for thumbnails.
Quality 85 is the default worth reaching for. Dropping from 100 to 85 typically removes 60β70% of the file with no perceptible change.
Important: lossy compression is one-way. Re-saving an already-compressed JPEG compounds the damage β each generation loses a little more, the way photocopying a photocopy does. Always re-export from the original, never from a compressed copy.
Lever three: the right format
- JPEG β photographs. Lossy, no transparency, universally supported.
- PNG β screenshots, logos, diagrams, anything with flat colour, sharp edges or transparency. Lossless, and terrible for photos, where it can be five times larger than an equivalent JPEG.
- WebP β typically 25β35% smaller than JPEG at matched quality, with transparency support. Every current browser handles it. The best default for the web today.
The most common format mistake is saving a photograph as PNG because "PNG is higher quality". It is lossless, which is not the same as better β it simply means it cannot throw anything away, which is exactly the wrong property for a photograph. The reverse mistake is saving a screenshot of text as JPEG, where lossy compression smears the letter edges.
Converting between formats never restores lost quality. A JPEG saved as PNG is just a bigger file containing the same already-degraded image.
What size does each platform want?
- Instagram post β 1080Γ1080 square, 1080Γ1350 portrait
- Instagram story / Reels β 1080Γ1920
- Facebook cover β 851Γ315
- X / Twitter in-post image β 1600Γ900
- LinkedIn banner β 1584Γ396
- YouTube thumbnail β 1280Γ720, under 2 MB
- Link preview (Open Graph) β 1200Γ630
- Full-width website hero β 1920 wide is plenty
- Email attachment β under 1 MB per image is courteous
Four mistakes that waste your effort
- Compressing before resizing. Resize first β it removes far more data, and compressing afterwards works on a smaller image.
- Re-saving JPEGs repeatedly. Each save degrades further. Keep the original and export fresh copies from it.
- Using PNG for photographs. Lossless is the wrong tool here; expect files several times larger for no visible gain.
- Uploading originals and letting the platform handle it. It will re-compress with its own settings, and you have no control over the result β often worse than if you had prepared the image yourself.
Do it
Image Resizer handles all three levers in one place: set the dimensions, pick the format, adjust quality, and it shows the before-and-after size before you download anything. It runs on a canvas in your browser, so the image is never uploaded β which matters for screenshots, documents and photos you would rather not hand to a stranger's server.
If you are preparing images for a site, Color Converter is useful for matching brand colours and checking contrast, and Base64 Encoder will inline a small image as a data URI. Everything else is on Pikkit, or under Web & Encoding.