How to Compress JPG to 100 KB, 50 KB or 20 KB Reliably

ilovejpg exists largely because of one recurring frustration: an online form rejects a perfectly good photograph because it is a few hundred kilobytes too large. If you need to compress JPG to 100 KB, 50 KB or even 20 KB, the task is entirely achievable, but it rewards understanding rather than repeated trial and error. File size is governed by two separate levers, pixel dimensions and compression quality, and most failed attempts come from pulling only one of them.

This guide explains why portals impose such limits, what a kilobyte actually means in this context, how the JPEG encoder decides how many bytes to spend, and a dependable order of operations that gets a file under the limit while keeping faces, text and signatures legible. It also covers the awkward cases, such as a scanned certificate that simply will not fit, and how to check the final file before you submit it.

Why application portals and forms set file size limits

Upload limits are rarely arbitrary. Examination boards, visa systems, recruitment platforms and university admissions portals may receive hundreds of thousands of submissions within a short window. Every applicant uploads a photograph, a signature and often several supporting documents, and each one must be stored, backed up, displayed to reviewers and sometimes printed onto admit cards or identity documents. Capping each image at a modest size keeps storage, bandwidth and page-load costs predictable.

There is also a practical display reason. A photograph that will appear on an admit card at roughly 3.5 × 4.5 cm needs only a few hundred pixels in each direction. A 12-megapixel phone image carries far more detail than the card can ever reproduce, so the portal asks for something smaller and saves everyone the waste.

• Storage and backup: millions of small files are far cheaper to keep than millions of large ones.
• Server load: smaller uploads complete faster and fail less often on slow or mobile connections.
• Consistency: a fixed size range nudges applicants towards similar dimensions, which makes reviewers' screens and printed documents look uniform.
• Security: strict limits reduce the scope for abusing an upload field with oversized or malformed files.

Many portals also impose a minimum size, for example a photograph between 20 KB and 50 KB. The minimum exists to stop applicants submitting tiny, heavily degraded images that cannot be used for identification. That matters later, because it means you should aim for a file comfortably inside the range rather than as small as possible.

KB versus KiB and why a 100 KB file can still be rejected

A kilobyte is not always the same number of bytes. Under the SI convention, 1 KB is 1,000 bytes. Under the binary convention, formally written KiB (kibibyte), 1 KiB is 1,024 bytes. The difference is only 2.4 per cent, but upload limits are often tight enough for that margin to matter.

Operating systems do not agree. Windows File Explorer displays sizes in binary units while labelling them KB, so a file it shows as 100 KB actually contains 102,400 bytes. macOS Finder has used decimal units since 2009, so the same file appears there as 102 KB. A portal's validation code might use either convention, and the help text rarely says which.

Treat every limit as decimal and aim a few per cent below it. A file of 95,000 bytes passes a 100 KB check under either convention, while a file of 101,000 bytes passes only one of them.

The same caution applies to minimums. If a portal demands at least 20 KB, a file of 20,100 bytes could be judged too small by a system counting in KiB. Aim for something like 25 KB in that case.

How pixel dimensions and quality each drive JPEG file size

The JPEG standard, published as ITU-T T.81 in 1992 and as ISO/IEC 10918-1 in 1994, compresses an image in a sequence of stages. Understanding those stages explains why the two levers behave so differently.

What the encoder does

• Colour conversion: RGB pixels are converted to YCbCr, separating brightness (Y) from two colour-difference channels (Cb and Cr).
• Chroma subsampling: because the eye is less sensitive to fine colour detail, the colour channels are usually stored at reduced resolution; the common 4:2:0 scheme keeps one colour sample for every 2 × 2 block of pixels.
• Block transform: each channel is divided into 8 × 8 pixel blocks and transformed with the discrete cosine transform (DCT), which expresses each block as a mix of smooth and fine-detail patterns.
• Quantisation: the DCT coefficients are divided by values from a quantisation table and rounded. This is the lossy step, and the quality setting controls how coarse those divisors are.
• Entropy coding: the remaining numbers, many of them now zero, are packed efficiently with Huffman coding.

Dimensions set the amount of work

The number of 8 × 8 blocks is proportional to the number of pixels. Halving both width and height leaves a quarter of the pixels and therefore roughly a quarter of the blocks to encode. That is why resizing is the most powerful lever available: going from 3000 × 4000 to 600 × 800 removes 96 per cent of the pixels before quality is even considered.

Quality sets the cost of each block

Lowering quality makes the quantisation coarser, so more coefficients round to zero and the Huffman coder has less to store. The relationship is not linear. Between quality 95 and 80 the file usually shrinks dramatically with little visible change; between 60 and 30 the savings become smaller while blocking and smearing become obvious.

Content sets the baseline

Two images of identical dimensions can differ greatly in size. Smooth backgrounds, soft lighting and plain clothing compress very well. Fine foliage, patterned fabric, sensor noise from a dim room and sharp text all generate high-frequency detail that survives quantisation and costs bytes. A photograph taken in good daylight against a plain wall is noticeably easier to compress than one taken under a kitchen light at night.

A reliable method: resize first, then reduce quality

The order matters. Reducing quality on a full-resolution phone photo to reach 50 KB would require such aggressive quantisation that the face would dissolve into blocks. Reducing dimensions to what the form actually needs first leaves the encoder with a small, manageable image that looks crisp even at moderate quality.

Prepare the image

• Read the portal's instructions for required pixel dimensions, aspect ratio, background colour and file size range, and note them before you start.
• Crop away unnecessary surroundings using your phone's or computer's built-in photo editor; ilovejpg does not include a crop tool, so do this beforehand if the framing needs adjusting.
• Open the cropped photograph in the ilovejpg JPG editor, which runs entirely in your browser so the image never leaves your device.

Resize to the dimensions the form needs

Use the resize control in pixels with the aspect-ratio lock enabled. If the portal specifies exact dimensions such as 200 × 230 pixels, enter those. If it gives none, choose a sensible size for the purpose: around 600 pixels on the long side is ample for an on-screen profile photograph, and 1200 to 1600 pixels suits a scanned document page that must remain readable.

Lower quality or use the target size option

Set the output format to JPG. You can then move the quality slider while watching the live size estimate, or enter the limit in the target size in KB field. The target size option searches for the highest quality setting whose output fits under the number you give, which removes the guesswork. Enter a value a little below the stated limit, such as 95 for a 100 KB portal.

Make small adjustments if needed

• If the estimated size is still too high at an acceptable quality, reduce the dimensions by another 10 to 20 per cent rather than pushing quality lower.
• For signatures and black-and-white documents, switch on grayscale; the colour channels then carry almost no information and compress to very little.
• A slight increase in contrast can make a faint signature or pale scan clearer at a lower quality, but avoid extreme values that clip detail.
• Download the result and confirm its size in your file manager before uploading.

Re-encoding in the browser also strips EXIF metadata, including any GPS coordinates and embedded thumbnail, which removes several kilobytes that contribute nothing to the form.

Typical dimension and quality combinations

The table below gives rough guidance for an ordinary, well-lit photograph saved as a baseline JPEG with 4:2:0 subsampling. These figures are approximate and are meant only as a starting point. Noisy low-light images, detailed backgrounds and textured clothing will produce larger files; plain backgrounds and soft lighting will produce smaller ones.

Use these values to decide where to begin, then let the live estimate or the target size option settle the exact quality. If a combination lands just above the limit, a modest reduction in dimensions is usually kinder to the image than a large drop in quality.

Signatures and passport-style photographs

Signatures

Signatures are among the easiest images to compress, provided they are captured well. Sign in dark ink, ideally black or deep blue, on clean white paper. Photograph or scan it in even light without shadows from your hand or phone, then crop tightly around the signature so that little blank paper remains.

• Turn on grayscale unless the portal explicitly asks for a coloured signature.
• Raise contrast slightly so the paper becomes close to pure white; uniform white areas compress almost to nothing.
• Keep the signature at a size where strokes remain smooth; 300 to 500 pixels wide is usually enough.
• Check the edges of the strokes at full zoom for blotchy blocks, which suggest the quality is too low.

Passport-style photographs

Application portals that request passport-style photographs usually specify a plain light background, a neutral expression, the full face visible from chin to crown, and no heavy shadows. Requirements differ between authorities, so follow the exact specification you have been given. From a compression point of view, a plain background and soft frontal lighting are your allies: they leave most 8 × 8 blocks smooth, so the encoder can spend its bytes on the face.

Avoid sharpening a passport-style photograph aggressively before compression. Sharpening adds high-frequency detail, which raises file size and can produce halos around the hair and jawline once quantised. Similarly, do not apply heavy saturation increases; skin tones that look natural are more acceptable to reviewers than vivid ones.

If a portal specifies dimensions in centimetres alongside pixels, trust the pixel values for the upload. Centimetre sizes describe the printed result, while the file itself is measured only in pixels.

When the limit cannot be met without visible damage

Occasionally a file refuses to fit. This usually happens with multi-page scans, dense certificates with small print, or very detailed photographs that must stay large. Forcing quality down to 20 or 30 in these cases produces blocky edges, colour smears around text and a halo of mosquito noise that can make a document unreadable.

• Reduce dimensions in smaller steps and recheck legibility at each one; text often remains readable at lower resolutions than people expect.
• Switch to grayscale for documents that do not need colour, which typically saves a substantial share of the size.
• Rescan in better light; a flatter, brighter scan with less paper texture compresses more efficiently than a shadowy phone photograph.
• Split a multi-page document if the portal allows several uploads, rather than squeezing every page into one file.
• Check whether the portal accepts PDF for documents; if it does, the ilovejpg JPG to PDF tool can combine appropriately sized page images into one file, though the PDF's size will still depend on the images inside it.
• Contact the organisation's help desk if a legitimately required document cannot be made legible within the limit, and keep a record of the attempt.

Avoid the temptation to compress the same JPEG repeatedly. Each save quantises the image again and compounds the damage, a process known as generation loss. Always work from the best original you have and make a single, deliberate export.

Checking the result before you submit

A file that passes the size check can still be rejected by a human reviewer if it is unclear. Before uploading, give the result a short inspection.

• Confirm the size in bytes, not only the rounded figure; on Windows the Properties dialogue shows both, and on macOS the Get Info panel does the same.
• Open the image at 100 per cent zoom and examine faces, eyes and text for blockiness or colour fringing.
• Check that the file extension is .jpg or .jpeg as required and that the image is the right way up.
• Verify the pixel dimensions match any stated requirement, since some portals reject images that are a few pixels off.
• Keep the original file and note the settings you used, so you can reproduce or adjust the result if the portal asks for a resubmission.

If the portal shows a preview after upload, look at it carefully. A preview that appears stretched or cropped suggests the aspect ratio does not match what the system expected.

Frequently Asked Questions

Conclusion

Getting a JPG under 100 KB, 50 KB or 20 KB is a matter of applying the right lever in the right order. Resize to the dimensions the form genuinely needs, then let quality, or a target size, finish the job. Aim slightly below the stated limit to cover the KB and KiB ambiguity, and stay comfortably above any minimum.

Above all, judge the result with your eyes as well as the number. A file that fits the limit but makes a face or signature unclear helps no one. Work from a good original, export once, inspect at full zoom, and keep your source file safe in case the portal asks you to try again.