Guide 7 min read ·

How to Reduce Image File Size

Four levers, applied in the right order, will cut most images by 80–95% with nothing visible lost.

Start with dimensions, not compression

This is the step almost everyone skips, and it is the one with the largest effect. File size scales with the number of pixels, so halving both width and height removes about 75% of the data before any encoder is involved. No compression setting comes close.

A 12-megapixel phone photo is around 4000×3000. If it will be displayed in a 800 px column, you are sending five times more pixels than any screen will use — including a retina screen, which wants 1600 px at most.

Sensible target widths

  • Full-width hero image: 1920 px (2560 px if you serve a 2× version).
  • In-article image or blog body: 1200 px.
  • Product or card thumbnail: 600 px.
  • Avatar or small icon: 200 px.
  • Email attachment: 1600 px is plenty for viewing on any screen.

Then pick the format that fits the content

Format choice is worth another large multiple, and it depends entirely on what the picture contains rather than on where it came from.

ContentBest formatWhy
Photograph for the webWebP (AVIF if it matters)25–60% smaller than JPEG
Photograph to send someoneJPEGOpens everywhere, no questions
Screenshot or UI capturePNG or lossless WebPSharp text, no ringing artefacts
Logo, icon, diagramPNG or SVGFlat colour compresses hugely
Anything needing transparencyPNG or WebPJPEG has no alpha channel

Then compress — and know where to stop

Only now does the quality slider matter. The relationship between quality and size is sharply non-linear: dropping from 100 to 90 typically halves the file with no visible change, while dropping from 60 to 50 saves little and starts to show.

  • Quality 90–100: for archives and further editing. Wasteful for the web.
  • Quality 80–85: the sweet spot for JPEG on the web. Loss is invisible in normal viewing.
  • Quality 70–80: fine for WebP, which holds up better at the same number.
  • Below 60: visible on photographs. Acceptable only for tiny thumbnails.

Judge by looking, not by the number. Compare the before and after at full size, and pay attention to the areas that break first: smooth skies, out-of-focus backgrounds, and edges between strongly contrasting colours.

Strip what you are not using

Camera photos carry EXIF metadata: model, lens, exposure settings, thumbnails, and often GPS coordinates. It is rarely more than 1–2% of the file, so this is a privacy step far more than a size one — but it is free, and publishing your home coordinates alongside a photo of your living room is a real mistake people make.

Any tool here that re-encodes an image already drops metadata as a side effect, because the output is built from raw pixels.

A worked example

  1. Start: 4032×3024 JPEG straight from a phone, 4.8 MB.
  2. Resize to 1600 px wide — now roughly 0.9 MB, and still sharper than any blog column needs.
  3. Convert to WebP at quality 80 — now roughly 180 KB.
  4. Metadata is dropped by the re-encode — GPS and camera details gone.
  5. Result: about 4% of the original, with no visible difference at display size.

The compression step gets the credit, but notice that resizing did the majority of the work.

When you have to hit an exact number

Upload forms with a hard limit — 2 MB, 500 KB, 200 KB — are a different problem. Rather than guessing at the slider, use the target-size option in our compressor: it searches quality, and then resolution if it must, to land under the limit, and tells you plainly if the target was not reachable.

Mistakes worth avoiding

  • Compressing an already-compressed file. Each JPEG round trip loses more detail for very little gain. Always work from the best original you have.
  • Compressing before resizing. You throw away quality, then throw away most of the pixels you paid for.
  • Using PNG for photographs. It is lossless, so it faithfully preserves a file five to ten times larger than it needs to be.
  • Trusting the number over your eyes. Different encoders mean different things by "80".
  • Discarding the original. Keep it; compression is one-way.

Related tools

Frequently asked questions

Can I reduce file size with no quality loss at all?

Only with lossless methods: resizing removes pixels but not fidelity at the new size, and lossless PNG or WebP re-encoding is pixel-identical. Any real reduction on a photograph involves lossy compression — the goal is loss you cannot see, not no loss.

What is a good file size for a web image?

Under 200 KB for in-article images and under 500 KB for a full-width hero is a reasonable target. What matters is the total page weight — a dozen 300 KB images is a slow page no matter how good each one looks.

Why did my file get bigger after compressing?

It was already optimised, or you converted to a less efficient format such as PNG. Our tools flag this and keep the original selected rather than presenting a larger file as an improvement.

Does compressing an image remove its metadata?

With these tools, yes — every output is re-encoded from raw pixels, so EXIF, GPS and camera fields are not carried across.

More guides