How to Reduce PDF Size for Online Applications (2 MB, 5 MB, 100 KB and Other Limits)
A PDF is too large for one of three reasons: it contains high-resolution scanned images, it embeds full font sets, or it carries hidden history from repeated editing. Fix the right one and a 12 MB scan becomes a 900 KB file that still reads cleanly on a consulate reviewer's screen.
The steps below use the browser-based PDF compressor, the page exporter for rescues, and the image compressor for rebuilds. Files never upload, which is why this method works for bank statements, medical reports and contracts.
Find out why your PDF is big before compressing it
Open the document properties (File - Properties in most readers) and look at the fonts list and page count. Then judge the content type: a document that was exported from Word is mostly text and font data; one that was scanned or photographed is a stack of images inside a PDF wrapper.
Text-based PDFs are usually 1-3 MB for a 20-page report and shrink only modestly, because the bytes are in embedded fonts. Scanned PDFs are 5-30 MB, and they respond dramatically to image downsampling — the single biggest size lever available.
Quick diagnostic: try selecting text with your mouse. If you cannot select a word, every page is a picture, and image settings are what you need to change.
Which compression level to pick for each limit
Use the compressor presets as a starting point, then judge the output on screen at 100 percent zoom before submitting.
| Target limit | Typical source | Preset | Extra step if still over |
|---|---|---|---|
| 5 MB (most job portals) | Scanned 10-20 pages | Medium | Downsample to 150 DPI, drop colour to greyscale |
| 2 MB (visa & admission forms) | Scanned certificates | Medium / High | Split into two uploads, or rebuild from 1200 px page images |
| 1 MB (government e-portals) | Mixed text + scans | High | Remove cover pages, compress the source images first |
| 500 KB (bank & KYC uploads) | Multi-page statements | High | Export pages to JPG, compress each to ~60 KB, rebuild |
| 100 KB (attachments, some forms) | 1 - 3 pages only | Maximum | Convert to greyscale, drop to 100 DPI, expect visible softening |
The 100 KB tier is the honest limit of the format: beyond three pages of a scan, quality degrades to the point where small print becomes unreliable. If the portal allows a larger ceiling, always take it.
The rebuild method: when compression is not enough
For a document that must stay readable at a brutal file cap, rebuild it from images rather than squeezing the existing PDF.
1. Export each page at a moderate resolution with the PDF to JPG tool — 1200 px on the long edge is enough for A4 at screen reading size.
2. Compress the page images to a fixed budget with the image compressor: quality 70 and greyscale where the source is black-and-white text. A clean A4 text page at 1200 px sits comfortably at 60-90 KB.
3. Recombine the pages in document order with the PDF merger. Because page images are already optimized, the resulting PDF is small and its text is crisp.
The trade-off is that a rebuilt PDF contains images only, so text is no longer selectable. Where the reviewer needs to copy text or search inside it, prefer lossless compression settings instead of rebuilding.
Keep text selectable: font and settings that matter
For text PDFs, the wins come from different places. Subsetting fonts (embedding only the glyphs actually used) rather than full families often removes a megabyte. Removing document history, comments, attachments and embedded thumbnails is another quiet gain.
Flattening form fields and annotations is usually safe; discarding layers (OCGs) and hidden optional content can be unsafe if the document had revisions you needed to keep. Never strip metadata that legally identifies a document, such as a signature block or a certification timestamp.
Merging several files under one combined limit
Portals that ask for "one PDF, max 2 MB" force you to combine CV, certificates and ID. Do it in this order: compress each source file to its own share of the budget first, then merge — merging first and compressing once afterwards gives a worse result because the compressor must apply one setting to documents of very different types.
Budget by page count, not equally by file: if your CV is three pages and your certificates are nine, the certificates get 75 percent of the allowance. Use the splitter to isolate any page that is disproportionately heavy (usually a colour scan of a seal or stamp) and compress just that page harder.
What to verify before you submit
Legibility at actual size: zoom to 100 percent and read the smallest print — stamps, watermark text, table footnotes and serial numbers are the first casualties of aggressive compression.
Page order and count: confirm no page was dropped during merge or rebuild.
Colour versus greyscale: if the document must show a coloured stamp, a signature in blue ink, or a photograph, do not let the preset convert it to greyscale.
File name: some portals reject spaces and non-ASCII characters in names; rename to something like application-john-doe.pdf.
Signature validity: a digitally signed PDF can lose its validity when reprocessed. If a signature must remain cryptographically valid, compress before signing, not after.
Frequently Asked Questions
How do I reduce a PDF to 2 MB for online applications?
Compress with the Medium preset first; if the document is a scan, that alone usually brings 10-20 pages under 2 MB. If it is still larger, downsample to 150 DPI, switch black-and-white pages to greyscale, and split out any single very heavy colour page to compress separately before merging again.
Can I compress a PDF to 100 KB without losing readability?
For one or two pages of plain text, yes. For multi-page scans, no — at that budget the images become too soft for small print and stamps. Either raise the target to 500 KB, reduce page count, or check whether the portal accepts multiple files instead of one.
Why did my PDF get bigger after compression?
The file was probably already optimized, and the recompression pass added a second generation of lossy image data while keeping the same overheads. Start from the original scan instead, and use a stronger downsampling setting rather than compressing an already-compressed output.
Is it safe to compress bank statements and contracts online?
Only when processing is local. Our compressor runs in your browser, so the document is never transmitted or stored on a server. Any service that uploads financial documents keeps a copy you cannot recall, which is a poor trade for a size reduction you can achieve offline.
Does compressing a PDF remove the text so it cannot be copied?
Only if you rebuild it from page images — that route produces a scan-like PDF without a text layer. Standard image downsampling and font subsetting keep selectable text intact.
Can I reduce PDF size on a phone?
Yes, the tools run in mobile browsers, so you can compress and merge a scanned document on the same device you photographed it with. Very large files may be slow on older handsets; split the document first and process it in halves if it stalls.
More Guides
How to Compress an Image to 100 KB (or Any Exact Size) Without Losing Quality
A tested method for hitting 100 KB, 50 KB, 20 KB or 500 KB exactly: dimension math, quality steps, and the formats that shrink best. Works on phone and desktop, no upload required.
Passport and Visa Photo Sizes by Country (US, UK, Schengen, India, Canada, Australia, China)
Exact photo sizes for US, UK, Schengen, India, Canada, Australia and China applications — millimetres, pixels, KB limits, background and head-height rules, plus how to prepare a compliant file.
JSON Formatter for API Debugging: Nine Common Errors and Their Fixes
Line and column errors explained: trailing commas, single quotes, unquoted keys, escaping, arrays versus objects, and the XML-to-JSON gotchas that break integrations. Paste and validate instantly.
How to Merge PDFs on a Phone for Free (Android and iPhone), No Upload
Scan pages, join them into one PDF and email it - all in the browser on your phone. Real page-count and size limits, the scan-then-merge workflow, and how to stay under a 2 MB portal cap without a desktop or an app.
QR Code Print Rules: Size, Error Correction and Contrast That Actually Scan
The numbers that decide whether a printed QR scans: module size, quiet zone, error-correction level (L/M/Q/H), matte vs gloss, print DPI and a real test routine. Generate a print-ready code free, in the browser.
Random Passwords vs Passphrases: Entropy Maths and Real Crack Times
How many bits a password really has, why length beats symbols, honest offline-crack-time maths, and where a password manager and 2FA matter. Generate a strong random password in your browser with nothing sent anywhere.
Word Count for Assignments: Limits, What Counts, and How to Trim Safely
Exactly what a word counter counts, whether the bibliography and headings are included, why 250 words is 250 words, and how to trim to the limit without breaking citations. Count free, in your browser, with no upload.
EMI vs Mortgage Loans: Tenure, Total Interest and the 28/36 Affordability Rule
The real EMI formula, why a longer tenure lowers the payment but raises total interest, the 28/36 affordability rule, part-payment and balance-transfer maths - all worked through. Calculate free, in your browser.
Base64 Data URIs for Images: the +33% Overhead and When Inlining Wins
Why base64 costs about 33% more bytes, how data URIs save or add requests, the HTTP/2 reality, cacheability trade-offs, and inline SVG vs bitmap. Encode and decode a data URI free, in your browser.
Metric and Imperial Conversions: the Factors, the Traps and the Rounding Rules
The exact factors for length, weight and temperature, the tonne vs ton vs metric-cwt trap, rounding rules for construction and cooking, and how to convert without introducing error. Convert free, in your browser.
Website Slow? Fixing LCP, Render-Blocking JS and Core Web Vitals
What LCP, INP and CLS actually measure, lab vs field data, the biggest real fixes - image sizing, render-blocking JS/CSS, caching - and how to check server latency. Be honest that a quick checker is not PageSpeed Insights.
How to Create and Submit an XML Sitemap (and What Actually Gets Pages Indexed)
A practical, honest walkthrough: build an XML sitemap, host it, reference it in robots.txt, and submit it in Google Search Console and Bing - plus why a sitemap is a hint, not a guarantee of indexing.
robots.txt Directives Explained: User-agent, Allow, Disallow, Crawl-delay and Sitemap
A clear, tested reference for every robots.txt line - how matching works, the $ and * rules, crawl-delay reality, and the mistakes that silently block the wrong things.
Meta Description Length: The Character Band and the Pixel Cut Behind the Snippet
There is no official character limit. Here is how Google actually truncates by pixel width, the ranges that usually survive on desktop and mobile, and how to write a description that gets shown instead of rewritten.
AI Crawlers, GPTBot and llms.txt: What They Mean for Your Site
Who the AI bots are (GPTBot, ClaudeBot, PerplexityBot, Google-Extended and more), how to allow or block each in robots.txt, and an honest take on what llms.txt does and does not do.
Domain Authority: What DA Actually Measures (and Why It Is Not a Google Score)
Domain Authority is a Moz prediction, not a Google metric. What DA and PA really mean, why they differ between tools, how to use them honestly, and how our checker fetches them.
How to Find Broken Links Before They Cost You Traffic and Trust
Why 404s and dead outbound links leak value, the HTTP codes behind them, a practical one-page check workflow, and how to fix each failure - using a browser-based link checker.
Google Index vs Crawl: Two Different Things Most People Conflate
Crawling, indexing and ranking are three separate stages. What crawled-but-not-indexed really means, which signals let a page be indexed, and how to verify status honestly in Search Console.
Keyword Density: Why There Is No "Safe" Percentage to Hit
What keyword density actually measures, why chasing a target percentage is a myth, when the number is a useful red flag, and how to use a density checker without stuffing.
UTM Parameters: A Clean Naming Convention and Reusable Template
What utm_source, medium, campaign, term and content actually capture, the rules for a consistent lowercase naming scheme, a copy-paste template, and the traps that wreck attribution.
HTTP Status Codes Every Site Owner Should Recognise
A plain-language map of 200, 301, 302, 404, 410, 403 and 5xx codes, what each does to your SEO, and how to read a URL's real status without guessing.
A GDPR-Ready Privacy Policy Checklist (and Why a Generator Is Only the Draft)
What a GDPR privacy policy must disclose, the many obligations that live outside the document, how CCPA differs, and how to use a policy generator without mistaking a draft for compliance.
Plagiarism vs Paraphrasing: An Honest Guide to Originality Checks
Why a browser plagiarism checker cannot compare your text to the internet, what matched-source detection actually needs, how to paraphrase legitimately, and how to use originality tools without fooling yourself.
Reading and Writing Regular Expressions Without Fear
A plain-language tour of regex syntax, the JavaScript flags (g, i, m, s, u), capture groups, the greedy-vs-lazy and engine gotchas, and how to test a pattern safely before you ship it.
CSV to JSON: How to Convert Cleanly (Delimiters, Headers and Quoting)
The traps that silently corrupt CSV data - quoted fields, embedded commas and newlines, header-vs-array shape, auto-typed numbers and BOM - and how to convert to JSON safely.
XML to JSON: What Gets Lost and How to Convert Faithfully
XML and JSON are not equivalent trees. How attributes, repeated elements, mixed content, CDATA and namespaces map across - and where silent data loss and array-vs-object ambiguity bite.
Minify, Compress or Cache? Three Speed Levers People Confuse as One
HTML minification shrinks bytes, compression shrinks transfer, and a CDN or cache cuts latency and rework. What each layer actually changes and the order to apply them.
Formatting SQL for Readable Code Review (Without Pretending It Validates It)
A SQL beautifier only reflows text - it cannot prove a query is correct. How to format for review, pick a house style, and combine it with real validation and EXPLAIN.
MD5 vs SHA-256 (and Why Neither Should Hash Your Passwords)
What cryptographic hashes are, why MD5 is broken and SHA-256 is not, why both are wrong for passwords, and why base64 is not hashing at all.
URL Encoding and Reserved Characters: What to Escape and When
Percent-encoding, the RFC 3986 reserved vs unreserved split, the difference between encoding one parameter value and a whole URL, and the double-encoding traps that break requests.
Favicon Sizes: The Complete, Honest List (and What Actually Matters)
A favicon is not one file. Which sizes serve which surfaces - tab, apple-touch, PWA 192/512, legacy ICO - what SVG can and cannot replace, and what a generator can really output.
Unix Time and the 2038 Problem: What a Timestamp Really Is
Epoch time is seconds since 1970 UTC ignoring leap seconds. Why 32-bit signed counters roll over in 2038, the seconds-versus-milliseconds bug, and always storing UTC.
UUID vs Auto-Increment IDs: Which Primary Key to Use
Random UUID v4 versus sequential integer IDs: index locality, key size, global uniqueness, enumeration risk, and the time-ordered middle ground (UUIDv7/Snowflake/ULID).
How to Upscale an Image Without Pixelation (Honest Limits)
What 2x/4x/8x interpolation really does, why enlarging cannot invent captured detail, smooth versus nearest-neighbour, and when upscaling helps versus when to leave the image alone.
WebP vs JPEG: When to Use Each (and When Neither)
WebP is 25-35% smaller and adds transparency and animation, but JPEG still wins on reach. When each fits, lossy versus lossless, and how compatibility and the quality slider change the choice.
PDF to Word: What 'Editable' Really Means (Digital vs Scanned)
Digital PDFs carry a real text layer and convert cleanly; scanned PDFs need OCR with imperfect results. What a converter can and cannot give you, and how to get a good edit.
How to Scan Documents to PDF on Your Phone (Free, Private)
Turn camera photos of pages into a proper PDF: shoot pages that scan well, drag to reorder, set page size, margins and quality, merge batches, and keep everything on-device.
BMI: What It Measures and Where It Is Wrong
BMI is weight over height squared - a cheap screening proxy, not body fat or a diagnosis. Where it misleads: athletes, children, the elderly, pregnancy, and different ethnicities.
How to Calculate Age for Documents (Exact Years, Months and Days)
Age 'as of' a reference date, completed years, leap-day (Feb 29) birthdays, and unambiguous date formats for admissions, exams, visas and benefits paperwork.
Counting Business Days: Weekdays, Deadlines and the Long-Weekend Trap
Working-day math for contracts and delivery SLAs: exclude weekends, remember holidays are not built in, and handle the inclusive-versus-exclusive start-date off-by-one errors.
What a Valid Invoice Needs (and What Depends on Your Country)
The elements almost every invoice carries versus the jurisdiction-specific extras - tax/GST/VAT number, tax breakdown, e-invoicing, sequential numbering. A template is not tax advice.
Text-to-Speech for Accessibility: What It Helps and Its Limits
Browser TTS speaks text with the voices already on the device. Who it helps (low vision, dyslexia, proofreading), where it falls short, and why it complements rather than replaces accessible design.
Downloading and Using YouTube Thumbnails: What's Fair
You can grab a video's preview image, but usage rights are separate from the download button. Ownership, fair use, embedding versus re-hosting, and the safe rules - plus how the downloader actually works.
How to Convert HEIC to JPG on Windows (Free, and Without Uploading Your Photos)
Open iPhone HEIC photos on a PC: use a browser HEIC converter that decodes locally, or install the HEVC/HEIF extensions. What HEIC is, why Windows struggles, and when to convert.