Image editors · Document and scan limit
Compress a scan to 100 KB and keep the text readable.
Document photos fail differently from portraits: small text turns to mush before anything else does. This page defaults to Sharp mode and tells you what it did, so an ID or receipt stays legible at 100 KB.
Why documents behave differently at 100 KB
A hundred kilobytes is 102,400 bytes. For a portrait, that is generous. For a photographed page of text it is tight, because the information that matters, the edges of small letters, is exactly what JPEG compression damages first. Lower the quality too far and letters grow fuzzy halos and blocks; shrink the pixels too far and the strokes of small print merge. A useful rule of thumb is that text stays readable on screen when the page is at least about 1,000 pixels across for A4 or Letter, and when ordinary body text is at least ten or so pixels tall.
That is why this page opens in Sharp mode. The tool will not let JPEG quality fall below 80% in that mode. It reduces the pixel dimensions first, keeping the aspect ratio, until the file fits, and tells you the final dimensions so you can judge whether the text is still readable.
Getting a better starting point
- Flat, even light. Shadows and gradients cost bytes and hide letters.
- Fill the frame. Crop so the document covers the picture. Empty table or floor space wastes pixels you do not have.
- Straight on. A page photographed at an angle needs more pixels to read. Use rotate and flip if it is sideways.
- A scanner or scanning app beats a phone camera for text, because it gives high contrast and a white background that compresses to almost nothing.
Shapes in the menu
The menu offers an A4 page at 100 dots per inch (827 by 1169 pixels; A4 is 210 by 297 mm), a US Letter page at 100 dpi (850 by 1100), and an ID card (85.6 by 54 mm, the ISO/IEC 7810 ID-1 size, at 200 dpi: 674 by 425 pixels). These are arithmetic from the paper size, not rules from any agency. Whether a portal wants those dimensions is something only the portal can tell you, so use the Width and Height boxes for its numbers. At 100 dpi, small print on a full A4 page is on the edge of legibility, so if the text is dense, try a larger limit or the exact-size card option for an ID only.
A worked example
Measured with this page's own code on a real photograph (a 3,840 by 2,160 pixel aerial image, 5.36 MB, so a picture rather than a page of text): Sharp mode with a 100 KB limit returned 99.7 KB at 1,155 by 650 pixels and quality 81%. Quality stayed at the Sharp floor and the pixel count fell instead, which is the behaviour you want for documents. For a photographed utility bill, press Resize without a preset and read the dimensions on the result line. If the account number is hard to read at that size, raise the limit to 200 KB, which gives the same page about 1.4 times as many pixels on each side, and compare.
Honest failure
If you fix an exact size, or choose "Keep pixels", the tool cannot shrink the dimensions. A busy full-page photo at 827 by 1169 can still exceed 100 KB at the lowest quality, and then no file is produced. The message names the smallest size reached. That is a better outcome than a file that quietly breaks your upload.
iPhone HEIC photos are accepted as input too. Browsers that cannot decode HEIC load a separate decoder (libheif 1.19.8, LGPL-3.0, unmodified) once, only after you choose a HEIC file; Safari uses its own.
What it does not do
It does not run text recognition, make a PDF or fix skew. It cannot guarantee that text is readable, only that the file is under the limit. Always zoom into the preview before you submit.
Frequently asked questions
Is my image uploaded to shrink it?
No. The photo is decoded, resized and re-encoded in this page on your device, inside a sandboxed frame whose security policy blocks network requests. Only you have the file.
Will the text in my document still be readable at 100 KB?
Often, if the scan is clean and the page is at least about a thousand pixels across, but there is no guarantee. Sharp mode keeps quality high and lowers pixels instead. Zoom into the preview and check small print before submitting.
What is the difference between Sharp and Balanced?
Sharp never goes below 80% JPEG quality and shrinks pixels more. Balanced goes down to 55% quality before shrinking pixels, which suits photos of people better than text.
Can I get a PDF under 100 KB from this?
This tool outputs a JPG. To wrap it in a PDF use the JPG to PDF tool, then check the PDF size, since the PDF adds a little overhead.
Why does a plain white scan compress so well?
JPEG stores flat areas in very few bytes. A clean black-on-white scan can fit in far fewer bytes than a photo of the same page with shadows and a coloured table.
Does the tool deskew or enhance the scan?
No. It only resizes and re-encodes. Straighten and crop first if needed.
Is metadata removed?
Yes. The picture is redrawn, so GPS, camera and date information from the original is not in the result.
Sources
Checked
- MDN: HTMLCanvasElement.toBlob() — how the canvas produces a JPEG at a chosen 0 to 1 quality value, which is the dial this tool searches (retrieved 2026-10)
- MDN: imageSmoothingQuality — the high-quality resampling setting used when pixels are reduced (retrieved 2026-10)
- ITU-T T.81 (JPEG) — the lossy DCT coding whose quantisation the quality value controls (retrieved 2026-10)
- MDN: createImageBitmap() — decodes the chosen file and applies its stored orientation (retrieved 2026-10)
- W3C: PNG Specification, Third Edition — PNG inputs are lossless; the output is always JPEG because only lossy JPEG has a quality dial (retrieved 2026-10)
- Apple: Capture formats on iPhone, iPad and Mac — iPhones save HEIF by default; Most Compatible saves JPEG (retrieved 2026-10)
- libheif (strukturag) — LGPL-3.0 HEIC decoder loaded only when a HEIC file is chosen and the browser cannot decode it (retrieved 2026-10)
Runs entirely in your browser. Your files are not uploaded, stored or sent anywhere.