Drop an image here, paste it with Ctrl+V or choose it
How to use
What it does
Resizes, converts and compresses images in your browser: the image is not uploaded to any server. Open it (drag it, paste it with Ctrl+V or choose it), adjust size, format and quality, and you will instantly see how it looks and how much it weighs.
iPhone HEIC photos
A HEIC photo (the iPhone's camera format) converts itself, in your browser, with heic-to (LGPL-3.0 licence): no need to change the phone's settings or share it another way first.
Compare before and after
In Compare, the original is to the left of the line and the result to the right. Drag the line to see where compression shows: it usually appears first on sharp edges and smooth gradients (the sky). The zoom buttons let you zoom in, in Compare too (that is how you see what “Upscale with AI” gains); in Original and Result you can also drag the image to move around it when it does not fit entirely.
Which format to choose
JPEG: photos. It loses some detail in exchange for weighing much less; no transparency. WebP: the same while weighing even less, and it does support transparency. PNG: lossless; ideal for screenshots, logos and text, but a photo in PNG weighs several times more than in JPEG. Quality (JPEG and WebP) is how much detail is discarded: between 75 and 85 it is usually not noticeable.
Convert to SVG (vectorize)
With SVG the image becomes shapes and curves, also in your browser. It works very well with logos, icons and drawings; a photo gives a large SVG that is no better than a JPEG. Next to the result you will see how closely it matches the original (the SVG is drawn again and compared pixel by pixel), how many paths and nodes it has and its size: raise “Remove specks” if you see little dots or want a lighter SVG.
Remove the background
Cuts out the subject and makes the background transparent, with an AI model (ormbg, Apache-2.0 licence) that runs in your browser: the image never leaves your device. The first time, the model is downloaded (about 45 MB) and kept. It takes a few seconds per image. To keep the transparency use PNG or WebP; with JPEG the background is filled with the colour you choose, and with SVG only the subject is vectorized.
Blur, cover or erase areas
Mark with a click whatever you want to cover —a face, a license plate, some data— and choose what to do with it: blur it, cover it with a color, or erase it for real (rebuilding what was behind it). The first two use an AI model (SlimSAM, Apache-2.0) to pick the area from the click; “Erase object” also uses LaMa (Apache-2.0) to fill it in, and is heavier and slower (a few seconds per click) — it works best on small or medium objects over a repeating background, not on a subject that takes up half the photo. The “Detect sensitive data” button reads the text in the image (OCR, built for screenshots) and marks only what LOOKS LIKE an email, phone number, card, ID number or key, with the same detectors the text Anonymizer uses — review the result, since it's neither always right nor foolproof. All of this happens in your browser: the image never leaves your device. It's applied before the rest of the adjustments, so it combines well with “Upscale with AI” or “Remove background”.
Product photo
Templates for selling: crops to the subject, centres it on a square canvas with a solid background and, if you want, adds a contact shadow. Amazon (2000×2000 px, pure white background, product at 85% of the frame) and Etsy (2000×2000 px square) follow their published requirements; Square (1080×1080 px) is generic, for social media or marketplaces with no size requirement of their own. Pairs well with “Remove background”: without it, it uses the original photo's background to find the subject.
Target size
Enter a size in KB (the one a form asks for, say) and the tool searches, trying qualities, for the HIGHEST one that doesn't go over it: no need to hand-tune the Quality slider. Only with JPEG and WebP, the formats with adjustable quality. If not even the lowest quality gets there, it says so instead of pretending it worked.
Auto-crop borders
Removes the single-colour, or transparent, border around the content: a product photo's white background, or the gap “Remove background” leaves around the subject. Nothing to select: if there's no uniform border, nothing changes. It runs before the size maximums, so “max width 800” is spent on what matters.
Colour palette
The most common colours in the image, with each one's share. Click a colour to copy its hex code.
Colour blindness simulator
The Colourblindness tab shows the image as someone with protanopia, deuteranopia, tritanopia or monochromacy would see it, without uploading anything: it's a colour filter applied by the browser itself.
Upscale with AI
Upscales a small image ×2 or ×4 with an AI model (Swin2SR, Apache-2.0 licence) that runs in your browser. It sharpens edges and textures without inventing detail: we measured it by shrinking photos to a quarter and upscaling them back, and they resemble the original more than with normal scaling. We also tried a model that gives “sharper” images and dropped it because it invented hair, eyebrows and lips. With a graphics card it takes seconds; without one, minutes, which is why it only upscales images up to 400 px per side (1024 px with a graphics card).
Metadata and privacy
A phone photo usually carries EXIF metadata inside: the camera, the date and sometimes the exact GPS location. The image you download from here carries none: it is redrawn pixel by pixel and only the pixels are saved.
Base64
Copy as Base64 gives the resulting image as a data: URL, ready to paste into HTML or CSS. Note: these are the bytes of the compressed file written as text, not the pixels. That is why it takes a third more than the file.
