How to Compress an Image to 100 KB (or Any Exact Size) Without Losing Quality

The three levers that control file size (use them in this order)

Every JPEG or WebP file is the product of three variables. Change them in the wrong order and you destroy quality for nothing.

1. Pixel dimensions. Halving width and height removes roughly 75 percent of the pixels to encode, and file size falls close to that. This is the only lever with real authority over large targets.

2. Quality (bitrate). The quantization setting. Between 85 and 75 percent most photographs lose 40-60 percent of bytes with no visible change; below 60 percent, blocky artefacts appear on skin and skies.

3. Format. WebP typically encodes the same photo 25-35 percent smaller than JPEG; HEIC smaller again but with patchy compatibility; AVIF much smaller but slow to encode and unsupported in some older browsers.

So: resize until you are within about 1.5x of the target, convert to WebP if the destination accepts it, then walk quality down in steps of five until you land just above the limit. Only after that should you revisit dimensions.

Start from the right source file

Screenshots and camera photos behave differently. A camera JPEG is already compressed once, so re-encoding it costs quality fast; a screenshot PNG is lossless but enormous, and converting it to JPEG can cut it by 90 percent with no visible loss because solid backgrounds and text edges survive chroma subsampling well.

Two rules before you touch the slider: never compress a file that has already been through three rounds of JPEG saving if the result must look clean (each generation adds artefacts), and if the image contains text, keep it PNG or WebP-lossless rather than JPEG — ringing around letterforms is what makes a compressed screenshot look bad.

If your source is a phone photo above 3000 pixels wide and the destination is a web page or a form upload, you almost never need more than 1600 pixels on the long edge. Decide that number first — it is the single biggest determinant of whether 100 KB is achievable.

Target-size cheat sheet: dimensions that make each limit realistic

The figures below are for typical photographs at quality 70-80 percent as JPEG. Text-heavy graphics, flat illustrations and low-noise images come in well under these; high-detail foliage and night scenes come in over.

If a form demands 20 KB and you send a 4000-pixel photo, no quality setting can rescue it — the maths is against you. Resize to the ranges above, then fine-tune quality.

Compress to 100 KB step by step (browser, no upload)

Open the image compressor and drop the file in. Nothing leaves your device, which matters for ID photos, receipts and unreleased product shots.

Step 1. Note the original dimensions and file size shown in the preview.

Step 2. Set the long edge to the value from the cheat sheet (about 1600 px for 100 KB). Keep the aspect-ratio lock on so the image is not stretched.

Step 3. Start at quality 80. Compress and read the output size.

Step 4. If the result is above 100 KB, drop quality in steps of five (80, 75, 70, 65). Stop as soon as you land between about 85 and 100 KB — leaving headroom is deliberate, because some portals round differently or re-encode on upload.

Step 5. Still above target at quality 65? That means pixel count is the constraint. Reduce the long edge by another 15-20 percent and repeat.

Step 6. Download. Check the file at 100 percent zoom on your own screen before submitting — you are the quality judge, not the slider.

When a portal still rejects the file

Rejections are rarely about size alone. Check them in this order:

Format whitelist. Many government and job-portal forms accept only JPEG. A WebP that is 40 KB still fails an upload that expects .jpg — convert with the format converter and rename deliberately.

Pixel bounds. Some forms specify both a file size and a resolution window (for example 600x600 to 1200x1200 px). Hitting the KB target while being outside that window still fails.

Aspect ratio. Identity-document uploaders frequently require a fixed ratio; a cropped-but-wrong-ratio image is auto-rejected. Use the cropper with the exact ratio before compressing.

Metadata. A few strict uploaders reject files with unusual EXIF blocks. Re-encoding through the compressor strips most of that as a side effect, which is why a "saved from the compressor" copy sometimes succeeds where the original failed.

Batch work: 50 product images at the same limit

Marketplace listings usually impose one shared ceiling (often 1 MB or 500 KB) across dozens of photos. Rather than repeating the process per file, fix the dimensions once and the quality once, then run the whole batch at those settings and spot-check the three largest outputs — the ones with the most detail — since they are the only plausible failures.

Order matters for batches: crop for the listing aspect ratio first, resize to the target long edge second, compress last. Doing it in reverse wastes effort on pixels you later throw away.

Quality versus size: what actually survives compression

Noise and texture cost the most bytes — sky gradients, skin, and dark cinematic shadows encode expensively. Flat colour, solid backgrounds and large uniform areas compress almost for free.

Practical consequence: a portrait against a plain wall can hit 50 KB looking flawless, while a forest scene at the same size looks muddy. For the forest, spend the bytes on dimensions instead of quality — a slightly smaller image at quality 80 nearly always beats a large one at quality 45.