Free Online Image Converter and Editing Tools

Why so many image formats?

Each one was designed for a problem the previous one handled badly, and none of them ever retired. JPEG arrived in 1992 for photographs and throws away detail the eye is poor at noticing, which is why it is small and why it cannot hold a transparent pixel. PNG answered that in 1996 with lossless compression and an alpha channel, at several times the size on a photograph. GIF is older than both and survives on one feature, animation, at a hard ceiling of 256 colours per frame. WebP and AVIF are the modern pair, both well below JPEG at matching quality, and both new enough that plenty of desktop software still refuses to open them. ICO is barely an image format at all: it is a container holding the same picture at several square sizes so a browser can pick the one it needs for a tab, a bookmark bar or a pinned shortcut.

So most conversions are a trade you should be able to name before you run one. Going PNG to WebP or through the JPG to WebP converter buys page weight and costs nothing a visitor will see. Going AVIF to PNG buys compatibility and costs file size, which is the right way round when an editor will not open the file at all. Animated WebP to GIF keeps the motion and cuts every frame to 256 colours, while pulling a still out of a GIF or flattening one to a photo format keeps the first frame and drops the rest. PNG to GIF is what people want when a chat client or a forum only accepts GIFs, and it is the trade with the least to gain, since a single-frame GIF is a worse PNG. And building an icon from a photograph works, though flat colour and heavy shapes survive 16px far better than a photograph does.

Metadata is the one that is worth doing first

A photograph taken on a phone typically carries EXIF metadata recording the camera model, its settings, the exact time, and unless location services were off, the GPS coordinates of where the shutter opened. Posting that file publicly publishes all of it. Most large social platforms strip metadata on upload, but a file attached to an email, uploaded to a forum, sent through a messenger that preserves originals, or hosted on your own site keeps everything.

How the stripping is done matters. The obvious method is to draw the image to a canvas and re-encode it, which does remove the metadata and also re-compresses every pixel, so a JPEG loses quality for a metadata edit. The EXIF remover instead removes the metadata segments from the container and leaves the compressed image data byte for byte identical, so the picture is unchanged. It also shows you what was in the file before you strip it, which is usually more interesting than people expect.

Converting is a second route to the same place, and a blunter one. The converter decodes the image, draws it onto a canvas and encodes it again, so EXIF, GPS and any colour profile are gone from the output, along with the original encoding. The exception is the rename: a JFIF renamed to JPG is a byte for byte copy, so everything in it, metadata included, comes through untouched. Treat losing metadata in a conversion as a side effect you happened to want, not as the way to do it.

What resizing and recompressing cost

JPEG and WebP are lossy: they discard information the eye is poor at noticing, and that information does not come back. Every save is a fresh loss, so an image opened, edited and saved repeatedly degrades visibly, and running it through JPG to PNG afterwards preserves the damage rather than repairing it. PNG is lossless and the right choice for screenshots, line art, logos and anything with sharp edges or transparency; for photographs it produces files several times larger than a JPEG nobody would tell apart.

WebP is the current default for the web, typically 25 to 35 percent smaller than JPEG at the same visual quality, with transparency support and universal browser support since 2020. The one thing to be careful about with the image converter is direction: converting to a lossless format never improves a lossy source, and converting a PNG screenshot to JPEG introduces ringing around text that no quality setting removes cleanly. JFIF to PNG is the same caution in a different costume, since the source is JPEG data whatever the extension says, and the PNG that comes out is lossless from that point forward rather than restored.

The two tools that are about the look

Cutting a subject out of its photo is now a job for the AI background remover, which sits with the AI tools because a segmentation model does the work: it recognises the person, pet or product and traces it, so a busy street or a cream wall separates as well as a studio sweep. It still runs client-side and writes the cutout at the resolution you gave it. Edges remain the hard part, since hair and fur are partly transparent, and the tool recovers the true colour of those pixels so no rim of the old background is left behind. Resize or compress the cutout here afterwards if the destination needs a smaller file.

The dither tool goes the other way, reducing an image to a small palette on purpose. Its six modes are genuinely different techniques rather than presets: Floyd-Steinberg distributes the full quantisation error across four neighbours and is the most photographically faithful; Atkinson, written by Bill Atkinson for the original Macintosh, passes on only three quarters of the error, which blows out highlights into the high-contrast MacPaint look; Bayer 4x4 uses an ordered threshold matrix for a regular crosshatch; the halftone mode samples along a rotated lattice the way a real halftone screen is angled on press; threshold is a hard one-bit cut at mid-grey with no dithering at all; and glitch keeps the colour, snaps it to the palette and scatters displaced blocks along one random band.