Image Optimizer — Compress to WebP & AVIF (Free, In-Browser)

The right quality is the lowest one where you cannot see the difference

Lossy compression works by discarding the detail your eye is least likely to miss. How much it can discard before you notice is not a property of the format — it is a property of your picture.

Foliage, a crowd, a textured wall: all busy already, so the encoder's simplifications hide inside noise that was there anyway. A flat sky gradient, a screenshot with text, a logo on white: nothing to hide behind, so the same setting shows as banding or fuzzy edges immediately.

Which is why a blanket claim like "this format is thirty per cent smaller" is describing somebody else's image. The number that means something is the one shown next to each file after processing: the bytes actually removed from this file.

Finding your setting

1. Start high — around 80 to 90 — and process one representative image.
2. Look at it at the size it will really be displayed.
3. Step the quality down and look again.
4. Stop one step before you can see a difference, then reuse that setting for images of the same kind.

The step people skip is the second one. Judging a web image zoomed to full screen makes you keep quality nobody will ever see, and that single habit accounts for a lot of unnecessarily heavy pages.

The formats

• WebP — the default. Broad support, keeps transparency, sensible for almost everything.
• AVIF — usually the smallest at a given quality, but not every browser can encode it. The tool detects that and tells you rather than failing silently.
• JPEG — no alpha channel, so transparency is flattened onto a white background before encoding. Fine for photographs, wrong for anything with transparency.

The biggest win is not compression at all

It is serving an image far larger than it is displayed, and it is the most common reason a page is slow. The reason it goes unnoticed is that the page looks perfect — nothing visually indicates the waste.

Take a photo uploaded at 4,000 pixels wide and displayed at 800. That is a fifth of the width, and because pixel count goes with area, it is a twenty-fifth of the pixels — 96% of the data is downloaded and thrown away before a single byte of compression saving is considered. No quality setting competes with that.

Use the max-width option here to downscale during the re-encode, or work out the dimensions first with the image resizer. Resize first, compress second; doing it the other way round wastes the compression pass.

What re-encoding costs you

All metadata goes: capture date, camera and lens, and any GPS coordinates. For a web upload that is a benefit — you were probably publishing your location without meaning to. For your archive copy it is a loss, so keep the original somewhere separate.

Never re-compress a compressed file

Every lossy pass throws detail away permanently, and the losses stack. Re-encoding an already-compressed JPEG at a lower quality does not undo the first pass, it adds a second — and the artefacts of the first become detail the second one faithfully preserves.

Always go back to the original. The tool accepts WebP and AVIF as inputs, which is convenient when the original is genuinely gone, but treat it as a last resort rather than a workflow.

Everything runs in your browser through the canvas API, so no image is uploaded, and the per-file saving is reported so you can judge the trade for the picture in front of you rather than from a general rule.