How to compress a PDF to 100 KB (without ruining it)

Before compressing anything, it helps to know whether your document is mostly text or mostly images, because the two need opposite treatments.

A PDF exported from a word processor or portal often contains selectable text and fonts, though it may also contain images. Some files contain extra metadata or redundant objects; a structural rewrite can help in some cases, but no specific saving or complete removal is guaranteed.

A scanned PDF often contains page images. Rebuilding pages at lower image quality or resolution can reduce size, but it can also damage text and remove any searchable text layer, links or interactive features. Check both the result and the portal’s requirements.

You can tell which you have in about five seconds: zoom into the text at 400%. If the letters stay sharp at any zoom level, the page contains real text. If they turn into soft pixels, you are looking at a scan.

Compression is not one operation. Applied in the right order, you often reach the target without any visible change at all.

• Start with structure-only compression when selectable text and document features matter. This site rewrites the PDF and resets common document properties; it does not promise to remove XMP, attachments or every hidden object. Savings vary, and a rewrite can invalidate signatures.
• Compare the before and after sizes. Some files are already compact and cannot be reduced by a structural rewrite. This tool keeps the original when a rewrite would make it larger.
• If the file remains too large and you can accept the trade-off, allow the tool to rebuild pages as JPEG images. Automatic mode may do this when the structural pass remains above target; rebuilt pages lose selectable text and may lose interactive features.
• Try a lower maximum render resolution if you need a smaller rebuilt file, but check small text, signatures and stamps carefully. Some targets cannot be met at a usable quality. The app reports the actual result rather than guaranteeing the requested size.

In this tool 100 KB means 102,400 bytes. Spread over ten pages, that is roughly 10 KB per page before PDF overhead. Whether a page remains readable at that budget depends on its content, resolution and encoding; do not assume a ten-page scan can fit legibly.

Small digits, Devanagari matras, signatures and stamps are especially important to check when reducing a scan. Colour can matter for official documents; do not convert to grayscale or alter a photo unless the recipient allows it. This PDF workbench does not offer a grayscale conversion control.

If a portal truly demands 100 KB for a twenty-page colour scan, consider whether the portal wants the whole document or just the relevant pages. Most application forms, on inspection, ask for an extract: the marksheet, not the entire transcript. Splitting the PDF first is the highest-leverage move available, and it is free.

Compression tools are happy to produce a file that passes the byte check and fails the human one. Before uploading, do three things:

Open the compressed file at 100% zoom and read the smallest text on the densest page. If you cannot read a serial number, the verification officer cannot either. Zoom to 400% on a page of body text — artefacts should be subtle noise around letter edges, not letters merging into each other.

Check that the page count is unchanged, and that the first and last pages are the ones you expect. A surprising number of compression problems are actually merge problems.

Look at the result card and the downloaded file size once more. A file below the maximum is normally sufficient if there is no minimum; an exact byte count is not guaranteed. Check the portal’s definitions of KB and MB, and leave a margin where possible.

Some documents cannot reach 100 KB at any quality that leaves them readable. A thirty-page colour scan with photographs, stamps and signatures on every page simply carries more information than a hundred kilobytes can hold, and no encoder can invent a way around that.

The order of escalation: read the portal requirement again, because many forms mean 100,000 bytes and others mean 102,400, and a file at 103 KB fails one and passes the other. Then narrow the document to the pages that are actually required and compress those alone. Then convert colour scans to grayscale where nothing depends on colour, which typically removes a third of the data. Only after all three should you consider dropping resolution.

Never accept a result where the text has stopped being legible. A submission refused for unreadable documents burns a re-submission window, and on some portals there is no second window at all.