Compress Image to Exact KB: How Target-Size Encoding Works

The form says the photo must be under 100KB. Yours is 104KB. You drag the quality slider down a little, export, and get 97KB, but the face looks soft. You nudge it back up, export again, and land on 103KB. Rejected again. Twenty minutes later you have six versions of the same photo and no idea which slider position gives the best quality that still fits.

That loop is the whole problem with compressing by hand. You need to compress image to exact KB, a hard number, using a control that was never designed to hit a number. Target-size encoding flips the job around: you name the size, and the tool searches for the highest quality that fits under it.

Compress image to exact KB: why the slider fails

A quality slider controls how aggressively the encoder discards detail, but it does not control file size directly, and the relationship is not linear. Dropping quality from 90 to 80 might cut a photo by 40 percent, while the same move on a textured image might cut it by 10 percent.

That unpredictability is why manual guessing eats an afternoon. Each export is a guess, each guess teaches you almost nothing about the next one, and the result you finally accept is rarely the best quality that fits. It is just the first one that happened to land under the cap. A tool that searches instead of guessing removes the loop entirely.

Compress image to exact KB with binary search

Target-size encoding treats quality as a search problem. The engine picks a quality level, encodes, and checks the size. Too big, it tries lower. Under the cap with room to spare, it tries higher, repeating until it finds the highest quality that fits, usually in six to eight encodes.

A worked example makes the search concrete. Take a portrait that starts at 1.8MB as a JPEG. The target is 100KB. The engine tries quality 90 first and gets 210KB, too big. It tries 70 and gets 88KB, under the cap but with room wasted. Then it splits the difference at 80 and gets 112KB, still too big, then 75 and gets 98KB. One more check at 77 gives 101KB, just over, so the search settles on 75 at 98KB. Every step is automatic, and the answer is the best quality the cap allows rather than the first one that happened to fit.

SnapShrink's target-KB compressor runs exactly this search in the browser: type 100 in the Target KB field, pick an output format, and the before and after sizes confirm the file came in under the cap. If quality alone cannot reach the cap, the engine gently downscales the image and searches again, rather than handing you a file that misses the limit.

When the cap is unreachable, and what to change first

Some images cannot reach a small cap at any quality without turning to mush. A detailed landscape at full camera resolution will not fit under 100KB and still look like itself. When that happens, change things in this order, because each step buys more quality than the one after it.

First, resize to the size the destination actually displays. A portal that shows a photo at 600 pixels wide does not need a 4000-pixel source, and resizing alone can cut the file by more than half before quality even enters the picture. Second, crop empty borders and background so the quality budget goes to faces and text instead of sky. Third, switch formats if the destination allows it: WebP usually reaches the same cap at visibly better quality than JPEG, while PNG is the wrong choice for photos under a cap because it stores pixels losslessly. Only after those three steps should you accept a lower quality, because by then you have removed waste instead of detail.

A good tool tells you when it has hit this wall instead of failing silently. If the smallest achievable size is still over the cap, that number is useful information: it tells you to resize or crop and try again, rather than exporting the same doomed photo a seventh time.

A checklist before you upload

Run through this short check before submitting, and the form stops bouncing your photo back. Size, format, dimensions, and a full-zoom look at faces and text cover nearly every rejection reason a portal gives. It takes two minutes and saves a second upload round.

1. Confirm the file is under the cap with margin to spare. Portals measure bytes, and 100.4KB rounds to a rejection.
2. Confirm the format matches what the portal accepts. A WebP under the cap still fails on a form that only lists JPG and PNG.
3. Confirm the dimensions meet the portal's minimum. Some forms reject images that are small enough but too low resolution, which is a different error with the same unhelpful message.
4. Open the final file at full zoom and check faces and any text. If either looks soft, resize or crop and run the search again rather than accepting the first version that fits.

One more habit pays off across every portal you will ever use: keep the original. Portals change their caps, a 100KB version made for one form is the wrong size for the next, and re-encoding from an already compressed file throws away detail twice. Archive the original, generate each capped version fresh from it, and the next form takes two minutes instead of twenty.

FAQ

Is a 100KB photo going to look bad?

It depends on the source. A modest portrait at display size usually looks fine under 100KB. A huge, detailed landscape will look softer, because you are removing most of its data. Resize first and the same cap buys visibly better quality.

Can I compress a batch to the same KB target?

Yes. Drop the batch, set the target once, and each file gets its own search, since a flat portrait and a busy landscape need different quality levels to hit the same cap. Download the results as a ZIP when the batch finishes.

What if my photo cannot reach the target at all?

Resize it to the display size, crop empty borders, and switch to WebP if the portal allows it, then run the search again. If the tool reports the smallest size it achieved, use that number to judge how much more you need to cut.

That 104KB photo from the start of this article takes about a minute now: type 100 into the target-KB compressor, let the search find the best quality that fits, check the face at full zoom, and upload. The slider loop is gone, and the version you submit is the best one the cap allows, not the first one that happened to squeak under it.