Questions

The work happens on your device instead of on our servers, so a picture costs us nothing to process. What is left, hosting the site and the models, is paid for by optional tips on Ko-fi (ko-fi.com/getpng). Nothing is locked behind a tip, and there are no ads.

They stay in this browser tab. Nothing is uploaded, logged or stored; close the tab and they are gone. You can check: after your first picture, switch your browser to offline and the tool still works.

The first time, your browser fetches the AI model from getpng.app and keeps it: about 44 MB on most laptops and phones, 115 MB on a computer whose graphics card runs the larger model. A quick first result appears within seconds while that happens, and sharpens once the model is ready. Every picture after that starts at once, even with no connection.

The model is from the same family the paid tools build on. Where getPNG differs is after the model: the edge is refined against your picture at its full size instead of being enlarged from a small mask, and that step is not locked behind an HD tier.

Yes, in a recent Chrome, Safari or Firefox. A device without WebGPU, or whose graphics card refuses the main model, gets a lighter model on the processor: coarser at the edges, still full resolution, and the stats bar under the picture says which one ran. A very large photo on a phone with little memory is processed at a reduced size, which the page also tells you about.

Edge shift grows or shrinks the whole outline by a few pixels, and Feather softens it. A brush for fixing one area by hand is planned; it is not in this version.

Yes. Pick a background colour, then choose JPG in the export box. A JPG has no transparency, so it always needs a colour behind the cutout.

Download the PNG and open it in getSVG, the sibling tool, which traces it into editable paths. That works best for logos, lettering and flat artwork, not for photographs.