A screenshot turns fuzzy when it is saved like a photograph

A creator saves every image as JPEG.

The file gets smaller.

For a photograph, that is often exactly right.

It is not the same thing as keeping the image intact.

JPEG was built for a picture whose detail changes gradually. Skin, sky, fabric, leaves, a room in natural light.

It works in small blocks, removes detail its model expects your eye to miss, and usually lets a group of pixels share their colour information while brightness stays sharp.

On a photograph, that trade is quiet.

On a black letter against a pale field, it has nowhere quiet to go.

There, the edge is the information.

So the edge rings. A flat colour block smears. A thin line ends up less certain than the thing it was drawn to describe.

JPEG isn’t failing at its job.

It has been given the wrong job.

The quality slider is what makes this easy to miss.

A scale from 0 to 100 turns every image into the same question: how much quality would you like to buy?

For a photograph, somewhere around 75 to 85 is the working range. Below it, blocking and ringing near hard lines start to show. Above it, you are mostly buying data you can’t see.

That is good advice about photographs.

It is not a reason to set a screenshot to 85 and consider the matter closed.

A screenshot isn’t asking for a little more JPEG.

It is asking for a format that doesn’t discard the edge in the first place.

Use PNG for screenshots, line art, flat colour, and any graphic whose hard boundary carries the instruction.

PNG is lossless. It predicts each pixel from its neighbours and compresses what is left over, without throwing image detail away.

Which means a PNG has no quality dial to rescue it, and none to ruin it.

It also means the file follows the image. A detailed photograph makes a large PNG. A screenshot full of repeated flat colour compresses beautifully.

A big photographic PNG is not proof that PNG is wasteful.

A small JPEG is not proof that JPEG is better.

They are answering different questions.

Transparency asks the clearest question of all.

JPEG has no alpha channel. It assumes a complete rectangle of colour and hands you one.

PNG carries full-precision alpha, so a soft shadow or an anti-aliased cutout can fade into whatever background receives it.

GIF has transparency too, but only the on-or-off kind: a pixel is fully visible or fully gone, and its palette is small.

That is why a GIF cutout arrives with a hard fringe on a background it was never made for.

So don’t flatten a transparent graphic into JPEG because JPEG is smaller.

Keep the alpha when the background is part of where the graphic has to travel.

WebP and AVIF make the choice less binary.

Both compress meaningfully harder than JPEG at matched quality, and both carry full-precision alpha, which JPEG never could.

AVIF goes further and costs more: encoding is slow, and older pipelines still need a fallback.

Either can be a good delivery copy for either kind of asset.

The decision underneath hasn’t changed. Does the mode you chose preserve the detail that mattered?

And don’t make the newest format a ritual.

Make it a delivery decision.

For a photograph, start with JPEG in that 75 to 85 range when broad decoding is the first need. Move to WebP or AVIF when the delivery path can spend the encode time.

For a screenshot, choose PNG when sharp edges, flat colour, or transparency have to survive. Lossless WebP does the same work where the publishing path prefers it.

For shapes and text that began life as shapes and text, consider SVG. It stores paths and fills rather than pixels, so it scales with how complicated the drawing is, not with how large you need it.

Then keep the clean source.

JPEG quantization is not something a later edit can retrieve. Crop a JPEG by a single pixel and save it again, and the block grid shifts underneath the picture, which can raise artifacts that weren’t there before.

A clean original lets every delivery file start from the same image, rather than from the damage the last delivery file already carried.

The smallest file can be a better photograph.

It can be a worse screenshot.

It can be a transparent graphic with the transparency quietly removed.

Name the asset before you name the format.

Then choose the compromise that lets its important detail arrive.

The smallest file is not the right file when the edge is the message.