How to get a photo under 100KB for an upload form

Open the image compressor, switch it to "To a size", type the number the form gave you, 100 KB for a photo or 20 KB for a signature, and press the button. It encodes the picture, measures the real bytes, and tries again until it finds the best quality that fits, then shows you the finished size before you download. If the number cannot be reached without the picture falling apart, it says so instead of handing you a file the form will reject, and the rest of this page is what to do then.

Why the photo is 4 MB to begin with

A phone camera records around twelve million pixels per shot, a grid something like 4000 across by 3000 down. Uncompressed, at three bytes per pixel, that is 36 MB. The camera's own JPEG encoder squeezes it at a high quality setting because the phone is trying to keep every detail you might zoom into later, and it lands at a few megabytes. The form is asking for a fortieth of that, and it is asking for a picture it will display at a few hundred pixels wide.

File size follows two things, how many pixels there are and how much detail the encoder is told to keep, and the form only cares about one of them.

What compression throws away

JPEG works on blocks of eight by eight pixels. For each block it keeps the broad tones and drops the fine variation an eye does not register, and the quality number decides how aggressively. Around 70, most people cannot tell. Somewhere below 40 the blocks start to show as a faint grid, edges grow halos, and skin goes waxy, and no lower setting recovers that. So quality has a floor, and it is not far below where a form's cap usually needs it to be for a full-size photo.

Pixels have far more headroom. Halving each edge of the picture keeps a quarter of the pixels, and the bytes fall by roughly the same factor without the quality number moving. A 4000-pixel-wide photo shown at 400 pixels in a form is carrying ten times the width it needs and a hundred times the pixels. Those are the bytes to give up first, because nobody was going to see them.

The compressor works in that order. It searches quality first, between 95 and 20, up to seven encodes. If no quality fits, it scales the picture down and searches again, up to three rounds, and the result line tells you when it did: "after scaling to 60% of the original width" rather than a silent change. A PNG asked for a size comes back as a JPG, because PNG has no quality dial to turn. And what you see on screen is the actual encoded file, so the size it reports is the size the form receives.

When 100 KB will not come

The tool reports the smallest it managed and stops. That happens when the picture has too much in it: a full frame with a wall, a doorway and a person, all of it detail the encoder has to spend bytes on. More compression will not get there. Less picture will.

Crop it first. A passport-style photo wants the head and shoulders, and a form that caps the file at 100 KB usually states a pixel size or a ratio as well. Crop to that framing, and everything outside it stops costing bytes. Then compress what is left. A cropped photo of a face on a plain background reaches 100 KB with quality to spare, where the same face in a busy room does not.

A signature at 20 KB is the same move, more so. Sign on white paper, photograph it, and crop tight to the ink. Nearly everything in the crop is flat white, which JPEG stores in almost no bytes at all, so the 20 KB is easy once the desk and the shadow of your hand are gone.

If the form states pixel dimensions, say 200 wide by 230 high, use the resizer after cropping. Hit the dimensions the form named and the byte limit usually takes care of itself, because a picture that small has nowhere to put 100 KB. The resizer fits the picture inside the numbers without stretching it, so crop to the right ratio first or the shorter edge will land under its number.

Compress once, from the original

Every JPEG save throws detail away, and the detail does not come back on the next save. A photo that has been through a messaging app, screenshotted, and compressed twice is already at the floor before you start, and the tool will tell you it cannot reach the target on a picture that should have reached it with ease. Go back to the copy from the camera and do the crop and the compression once, from there.

The same applies in the other direction. WhatsApp re-encodes anything sent as a photo. If the point is to get a small, sharp file to someone, make it here and send it as a document, so the bytes that arrive are the bytes you made.

Where the photo goes

Nowhere. The encode runs in your browser with the codec it already has, the picture is read from your disk into the tab, and the file you download is built there. The images that fail these forms are passport photos, ID scans and signatures, which are the last things worth handing to a stranger's server to shave off a few hundred kilobytes. How that works, and how to check it, takes thirty seconds and needs no tools.

For a PDF with a size cap rather than a photo, the same reasoning about resolution before quality is measured on three files here. Corrections to support@filetity.com.