Toolpit
Loading your page…
Toolpit
Loading your page…
The everyday jobs — resizing, cropping, converting between formats, in bulk — alongside four AI tools that remove backgrounds, erase objects, strip watermarks and enlarge photos. The AI models are quantized to download once and then run entirely on your own hardware, which is the whole point: those are exactly the tools that normally demand an upload, an account and a watermark.
Private — processed on your device, never uploaded
A grid of names does not help much when two of them sound alike. Find the row that matches what you are actually trying to do.
A cut-out with no background
Background remover — A segmentation model finds the subject. Export as PNG or WebP to keep transparency — JPG cannot store it.
A picture that is too small for how it will be used
Image upscaler — Reconstructs detail rather than smearing pixels the way plain resizing does.
Something in the frame that should not be
Object removal — Brush over it and the gap is filled from the surrounding pixels.
An exact pixel size, or a whole folder at one size
Resize image — Straight scaling, in bulk. Use the upscaler instead if you are going significantly larger.
The wrong file format
Convert image — Moves between PNG, JPG, WebP and AVIF. See the format notes below before choosing.
The right size but the wrong framing
Crop image — Trims to a ratio or an exact box. Cropping discards pixels; resizing keeps them all.
JPG is for photographs and nothing else: it throws away detail your eye is unlikely to miss, which makes it small, and it cannot store transparency. PNG is lossless and handles transparency, which makes it right for logos, screenshots, diagrams and anything with flat colour or hard edges — and needlessly large for a photograph. WebP does both jobs and lands roughly 25-35% smaller than the equivalent JPG or PNG, with support in every current browser. AVIF is smaller again, sometimes dramatically so, at the cost of slower encoding and patchier support in older software. For the web, WebP is the safe default and AVIF the aggressive one. For a file someone else will open in an unknown program, PNG or JPG remains the least surprising choice.
Making an image larger means inventing pixels that were never captured. Ordinary resizing does this by averaging neighbours, which is why an enlarged photo looks soft — no new detail appears, the existing detail is simply spread over more pixels. An upscaling model does something different: it has seen a great many images and predicts what plausibly belongs in the gaps, so edges stay sharp and texture survives. It works well on photographs and on artwork with clean lines. It cannot recover something that was never in the file — a face that is thirty pixels across has no expression to restore, and no model can put one back.
Resizing keeps the whole picture and changes how many pixels describe it. Cropping keeps the pixels and throws away part of the picture. The distinction matters when something requires an exact aspect ratio: a square profile photo from a landscape original has to lose the sides, and no amount of resizing will do it without squashing faces. Do the crop first, to fix the composition, then resize to the pixel dimensions you need — that order avoids scaling detail you are about to discard.
No. The model file is downloaded to your browser the first time you use one of those tools and cached for later visits, and it then runs on your own device. The download is the model coming to you, not your image going out — open your browser’s network panel while a tool runs and you will not see your file in any request.
It downloads the model, which is tens of megabytes, and that happens once per browser. After that the model is served from cache and the only cost is the processing itself, which depends on your device and the size of the image.
Yes. JPG is lossy, and each save discards a little more, so repeatedly editing and re-saving a JPG degrades it in a way that cannot be undone. Keep an original in PNG or in the format your camera produced, and export to JPG as the last step rather than as your working format.
Resize, convert and the watermark remover all take as many files as you like and apply the same settings across the set, then hand back a ZIP. The whole set saves as one download — 10 tokens, not 10 per file — and there is no upload queue, so the practical limit is your device’s memory.
Private by design — your files are processed on your device, not uploaded. Every tool on this page processes files in your browser, with JavaScript and WebAssembly; your files are not uploaded to or stored on our servers.