Edit an image with layers

No, and this is the main reason the page exists. Doing five things across five tools means five decode-and-re-encode cycles, and every lossy save throws away a little more of the picture - so the fifth result is visibly worse than the first even though nothing you asked for was destructive. Here the whole document is one description of what you want, compiled into a single pass: one decode, one encode, whether you made three changes or a hundred. The original bytes are only read once and the output is only written once.

No. Every layer, every filter and the final export are all computed by code running in this tab, and a content security policy the browser enforces stops this page sending your image to another domain. There is no account, no project stored on a server, and nothing recoverable by us - which also means a draft lives only in this browser, on this device.

Enough for any real page layout, and the limit is your device rather than a number we picked. Each image layer holds a decoded copy in memory, so a dozen large photographs is a real amount of memory and the page checks before it accepts one. Text and shapes are geometry and cost almost nothing - a hundred of those is not a problem.

Hiding removes it from the export; locking protects it from being edited but leaves it in the picture. They are kept apart deliberately, because conflating them is how you lock a layer to keep it safe and then find it missing from the file you downloaded.

Yes. Instead of picking a format and quality, you can give the export a requirement - a maximum size in kilobytes, maximum dimensions, or a destination like a social platform - and the same solver the constraint pages use will work out the quality and dimensions that meet it, then measure the finished file to check that it actually did. It reports the size it reached rather than the size it aimed for.

Three: a sans-serif, a serif and a monospace, each resolving to whatever your device provides. That is a smaller list than most editors offer and it is the honest one - the text is rendered by a background thread that cannot see web fonts, so naming a font we cannot guarantee would mean the exported image quietly differs from what you saw on screen.

No, and it is a refusal rather than an oversight. An SVG is a document that can carry scripts, external references and embedded content, so treating one as an image means either rendering active content or writing a sanitiser and hoping it is complete. Neither is a promise worth making on a page whose whole point is that your file never leaves your device. Export an SVG to PNG in the application that made it and open that instead - and if a file is refused here, that is why, rather than something having gone wrong.

No. Fetching a URL on your behalf means this page making requests to servers you named, which is a category of thing a privacy-first tool should be slow to add - it can be used to probe networks the browser can reach and it puts our code between you and somebody else's server. Save the image first and open the file. If that changes, it will be because there is a use for it that a download does not already cover.

Yes. A redaction layer is not a black rectangle drawn on top: the pixels underneath are overwritten before the file is written, so the information is not in the download. This is the same operation the dedicated redaction page uses, and it is tested by decoding the exported file and measuring that the marked area contains no variation at all.