Images · 10 tools
Work on a picture, not a document
Five things people need from a picture and cannot do without handing it to somebody: make it a different size, cut it down to the part that matters, change a format their computer refuses, and use the words in it when the words are pixels. None of it has anything to do with documents, and none of it needs a server.
The tools
- Convert HEIC HEIC to JPG or PNG Open an iPhone photo anywhere
- Resize an image Picture to chosen size To a width you actually name
- PNG to JPG PNG to JPG And what happens to the transparency
- WebP to PNG WebP to PNG Transparency kept, nothing uploaded
- WebP to JPG WebP to JPG Choose the quality, see the size
- PNG to WebP PNG to WebP Smaller for the web, transparency kept
- SVG to PNG SVG to PNG or JPG At the size you choose, drawn here
- Crop an image Picture to part of it Drag across what you want to keep
- Image to text Picture to text Read the words out of a photograph
- Image to Excel Table photo to XLSX A photographed table, as a spreadsheet
Why a PDF site has image tools
Because the promise was never PDF, it was that your file stays on your device. The photograph somebody cannot open is usually a photograph of paperwork, the screenshot being resized to fit a form is usually of something private, and the picture whose text somebody needs is usually a receipt or a letter. Sending any of them to a stranger's server to fix a size or a format is the trade this site exists to refuse.
The decoding belongs to your device
There is no HEIC decoder in this bundle and there is not going to be one: a WebAssembly build of libheif is well over a megabyte, downloaded by everybody to serve the minority whose browser cannot already do the job. The browser's own codecs do the reading, which means HEIC works on iPhone, iPad and current Safari and generally does not on Windows. The tool says so, and names the device that can, rather than failing into a spinner.
Everything that is not the picture is left behind
A phone writes far more into a photograph than the photograph: the camera and the lens, the exposure, the date and time to the second, an editing history, and, with location switched on, the position the shutter was pressed at. Resizing, cropping and converting a format all remove those blocks from the file they hand back, and then measure the result — you are shown how many bytes of it were in your file and how many are in the finished one. The removal is done deliberately rather than assumed, and that distinction was expensive to learn: browsers do not agree. Chrome and Firefox write a re-drawn picture with nothing attached, and Safari copies the camera and rights blocks straight through into the new file — which is to say the assumption fails on the iPhone, the device whose photographs are the likeliest to carry a location in the first place.
Reading text out of a picture is a guess, and it is shown as one
Optical character recognition does not read. It measures shapes and proposes the characters most likely to have made them, which is why it is excellent on a flat screenshot of clean type and poor on a photograph taken at an angle in bad light, and why it has not been promised on handwriting, where it was never measured. The reader here scores each line and lists the ones it was least sure of, rather than handing back a tidy block of text that looks equally certain throughout — a wrong digit in an account number is the kind of error that reads perfectly and costs the most.
What is not here yet
Three things that were on this list are here now: resizing to a width you name, cropping, and PNG to JPEG. Stripping camera data did not become a tool of its own, because it is not a job — it is what happens to every picture those three redraw, and each of them measures it rather than claiming it. Still missing: rotating and straightening a photograph, and one instruction applied to a batch of pictures. Each arrives when it works, and not as a menu entry before then.