Skip to main content
GARGLOO

Guide · 5 min read

How to Compress a PDF Without Losing Quality

Shrink oversized PDFs for email and uploads: what actually makes PDFs big, which compression to pick, and how to check the result.

Last updated: 2026-10-07

Tools for this guide

The short answer

Use Smart mode and judge by zoom, not by bytes: text stays vector-sharp at any size while the embedded images do the shrinking, so check fine print at 150–200% zoom — if it reads clean there, the quality survived.

Lossless mode changes nothing visible, and Maximum mode turns every page into a picture for the smallest file at the cost of selectable text. When a target size would force visible damage, the tool says the file cannot fit instead of silently ruining it.

Find out what's making it big

A 2-page text PDF should be kilobytes. If yours is tens of megabytes, open it and zoom to 400%: fuzzy edges mean scanned images, which compress well; crisp text means embedded high-resolution assets or uncompressed streams.

On Gargloo, drop the file into the Compress PDF tool — it reports exact before/after sizes per preset, so you can diagnose without guessing.

Pick the right strength

Start with the lossless re-save: it removes unused overhead without touching content. It only helps files that actually have overhead, but it never degrades anything.

If that's not enough, use image-based compression at a balanced preset (around 150 DPI equivalent). Reserve the strongest preset for scans you only need to read on screen.

Check the result before sending

Compare file size against your email or portal limit with margin to spare. Then spot-check 2–3 pages at 150–200% zoom: body text should stay crisp, and fine print like footnotes should remain legible.

If quality suffered, re-run one preset gentler rather than re-compressing the already-compressed file — repeated compression stacks artifacts.

When compression isn't the answer

If only a few pages are heavy, split the document and compress just those, then merge it back. If the PDF contains full-page photos you don't need, export those pages as images first.

Password-protected or DRM-restricted PDFs must be unlocked by their owner first — no honest tool can or should bypass that.

How Smart mode preserves text while shrinking images

Smart mode, the default in the Compress PDF tool, treats a PDF as two separate layers with different rules. Embedded raster pictures — scanned sheets, photographs, phone captures — are decoded in a Web Worker, downscaled and re-encoded as compact JPEG data, then placed back exactly where they came from. Text characters, fonts, vector drawings, margins, and layout instructions pass through untouched, so headings hold their weight, paragraphs keep their line breaks, and tables stay aligned. That separation is why Smart output looks identical in structure while weighing far less in bytes.

The result note after each run describes what happened in plain language: how many embedded images were recompressed and whether all text stayed selectable. Because text operators survive the round trip, selection, copying, searching, and screen-reader access continue to work exactly as before. Measured production runs confirm the preservation directly, with extracted word counts identical before and after on every fixture containing text. When a document carries no embedded images at all, Smart mode says so openly instead of fabricating savings, which points you toward removing pages rather than twisting dials.

Use Lossless first and read its notice honestly

Lossless mode re-saves the document with object streams and removes removable overhead without altering a single pixel or character. It never reduces quality under any circumstance, which makes it the correct opening move on files of unknown origin: exported reports, generated invoices, documents assembled from several sources, or anything that may carry duplicate objects and uncompressed streams. The run completes quickly, and the before-and-after sizes tell you immediately whether overhead was the problem.

When Lossless cannot help, the notice states it plainly: the file has no removable overhead and Smart mode is the next step. Treat that message as a diagnosis rather than a failure, because it rules out an entire category of causes in seconds. Files that respond to Lossless are typically text-native documents with structural waste; files dominated by scanned images or photographs barely move, since their weight lives in pixels rather than container overhead. Either outcome sharpens the next decision instead of leaving you to guess.

Aim at a size with target presets instead of guessing

Once you know the cap stated on the upload form, declare it as a target instead of compressing blindly and checking afterward. The Compress PDF tool offers preset chips at 100KB, 200KB, 500KB, and 1MB alongside a custom field where you can type a value just under the stated limit with margin to spare. The target drives Smart and Maximum modes through their quality ladders automatically, stopping at the mildest pass that lands under the goal. That automation replaces the old loop of compress, check, repeat with a single stated intent.

The interface is explicit about one boundary: Lossless re-saving cannot aim at a size, since it performs a fixed structural cleanup rather than a quality tradeoff. When the engine reaches the goal, the result note confirms it; when even the strongest honest pass lands above the goal, the note reports the closest achievable size instead of silently destroying readability to hit the number. That honest floor protects documents from the classic failure of tools that claim to fit any file under any cap. Set the target slightly below the real cap so portal overhead cannot push a borderline file over.

Understand what scanned, photo, and text pages save

Measured Smart-mode results on synthetic fixtures, reproduced end to end in headless Chromium against the production build, let you forecast behavior before running. Scanned documents saved 58.6%, photo-heavy pages saved 77.8%, mixed text-and-image pages saved 71.4%, and a single large photo page saved 96.7%, all in 0.3 to 2.1 s with text extraction identical before and after wherever text existed. Scans and photographs therefore respond strongly in about a second, because their pixels carry the redundancy that recompression removes. Mixed reports with diagrams and snapshots land in the middle, with savings proportional to their image share.

The same synthetic-fixture measurements explain the flat cases without apology. Text-only pages saved 0% and returned byte-identical in 0.3 s, as did an already-compressed tiny file, because no embedded images existed to recompress and the tool issued an under-5% notice rather than pretending. Diagnose with that split in mind: zoom pages to high magnification and look for paper texture and fuzzy edges that mark scans, or run Smart once and read the note. Image weight means compression solves the problem; text weight means the document needs fewer pages, leaner source exports, or downsampled screenshots before exporting.

Handle password, damage, and size-cap rejections

The Compress PDF tool accepts files up to the 150MB cap and processes everything on your device, which sets firm boundaries worth knowing before you start. Files above 150MB are refused at selection time regardless of mode or target, so oversized originals must be reduced where they were created: export in parts, lower scan resolution at the source, or divide the document before bringing it here. Password-protected files are refused with a clear message asking you to remove the password first by re-saving without one, then retry. Damaged files receive a message suggesting a re-save from the original application. Neither case can be solved with stronger settings.

Rejections after successful compression usually involve rules beyond bytes, so verify the surrounding details in order. Confirm the extension is exactly .pdf and the filename uses plain characters, since strict validators stumble on unusual symbols and doubled extensions. Confirm the compressed size sits below the stated cap with margin, because portals often measure decoded size or add handling overhead. Confirm the file opens locally without prompts. When every check passes and uploads still fail, re-download the compressed copy fresh and retry, since interrupted downloads truncate bytes and read as corruption.

Verify text selection and layout after each run

Every compression run ends with exact before-and-after sizes, so compare the output against the margin-adjusted target before anything else. Then inspect content two ways: visually, by opening two or three representative pages at 150 to 200% zoom and reading body copy, footnotes, figure captions, signatures, and fine print; and structurally, by dragging across a paragraph to confirm it highlights and pastes cleanly elsewhere. Smart mode preserves layout identically and keeps text selectable, while Maximum mode states openly that text becomes non-selectable images. The selection test distinguishes the two outcomes in seconds.

Fold verification into a repeatable routine that survives busy days. Check the page count against the original so no sheet went missing, confirm orientation and margins match, and review any pages with dense tables or small labels where artifacts appear first. Never stack passes by re-compressing an output when the original remains available; repeated encoding compounds marks on images while text pages stay the same size. Archive the original, submit the verified copy, and keep both until acceptance is confirmed so follow-up requests for full quality can be answered immediately.

Split, extract, and merge when pages differ in weight

Long documents rarely weigh evenly: a text-native report may hide three scanned appendices, or a photo section may dwarf fifty pages of prose. When only some sheets are heavy, restructure instead of squeezing the whole file. The Split PDF tool divides a document into ranges, while Extract PDF Pages pulls specific sheets or ranges such as 2, 5 to 8, and 12 into one new file. Both operations rearrange original content without re-rendering it, so extracted sheets keep full quality. Compress the heavy subset with a declared target, verify it, then combine parts with Merge PDF in the intended order.

This surgical sequence protects quality where blanket compression cannot. Text chapters avoid any encoding pass they never needed, while scanned exhibits receive the attention they deserve. Typical patterns include extracting required pages for a form and leaving appendices out, dividing a submission across several upload slots each with its own cap, or isolating one oversized photograph for separate handling. Name each part clearly so its purpose survives the recipient folder, and retain the untouched original until the reassembled delivery is accepted. Structure first, compression second, verification always.

Keep masters and generate delivery copies from sources

Repeated submissions punish anyone who re-compresses yesterday compressed copy for today portal. Maintain a masters folder holding the uncompressed original and its sources — exported text, scanned certificates, photographed exhibits — and treat every compressed upload as a disposable copy generated from those masters. When a new form states a different cap, return to the master and run one fresh pass with the new target rather than squeezing the previous output further. Regeneration costs seconds, with measured Smart runs finishing in 0.3 to 2.1 s, and always outclasses polishing an already-processed file.

The archive also answers the reversed request that follows many approvals: a print-quality packet for meetings, binders, or on-site review. Screen-optimized uploads and printed handouts are different products, and the master serves the latter without reconstruction. Name masters distinctly from deliveries so a tired evening never sends the full-resolution original through a small capped slot, and record which cap each delivery was tuned against. One honest pass from current sources, verified at reading zoom with selection confirmed, remains the entire quality system; everything else is filing.

Compare Smart and Maximum outputs side by side

When both modes seem plausible for a stubborn file, run a deliberate comparison from the untouched original rather than guessing. Compress once in Smart mode with the declared target and record the output size plus the result note; then return to the original and run Maximum mode the same way. Open both outputs beside each other at 150 to 200 percent zoom across representative pages — body copy, footnotes, diagrams, signatures, stamps — and note where legibility diverges. Perform the selection test in each file by dragging across a paragraph: Smart output highlights and pastes cleanly, while Maximum output presents pages as pictures exactly as the interface warns.

Choose with the recipient workflow in mind rather than the smaller number alone. Filings that parse, search, quote, or screen-read document text need Smart output even when Maximum shaves further kilobytes, since flattened pages frustrate every downstream reader. Maximum earns its place for scan-only bundles consumed purely on screen, where no text layer existed to preserve and the smallest file genuinely serves the viewer. Submit only the compared winner, archive the original alongside both trials, and note the decision rationale so future submissions to the same destination start from evidence instead of repeating the experiment.

Handle very large files near the 150MB ceiling

Documents approaching the 150MB acceptance cap demand staging discipline that smaller files forgive skipping. Close competing tabs before compressing anything above a few dozen megabytes so the browser holds maximum memory for decoding, since all processing happens in the tab rather than on a server. Run one mode at a time from the original — Lossless first for structural savings, then Smart with the declared target — reading each result note before proceeding instead of stacking passes. Measured Smart runs finished even a 50 MB photo page in 2.1 s with sequential page handling and canvases released between sheets, so patience is measured in seconds rather than minutes.

When a file exceeds the ceiling outright, reduce it where it was created before bringing it here: export long reports in parts, rescan at moderate resolution instead of maximum, or divide chapters into separate documents. Alternatively split the oversized file elsewhere first and compress each part independently, since smaller inputs process faster and a text-only part reports its honest 0% with an under-5% notice in 0.3 s rather than consuming a long render for nothing. Keep every original until compressed parts verify cleanly, because re-running from sources costs little while reconstructing deleted originals is impossible.

Document settings for recurring submissions

Monthly reports, compliance filings, and repeated applications punish anyone who rediscovers settings from scratch each cycle. Maintain a simple log per destination recording the stated cap, the mode and target used, the resulting size, and quirks encountered — validators rejecting long filenames, portals measuring decoded bytes above nominal caps, slots accepting multiple files. The next submission starts from a proven recipe instead of a fresh experiment, and Smart mode speed means even a from-scratch pass finishes in 0.3 to 2.1 s. What the log truly saves is diagnosis time around failures, because the quirks column remembers what frustration forgets.

Fold the log into archiving habits by storing each submitted copy beside its log entry under matching names, so resend requests months later need no regeneration or guesswork about which variant cleared validation. Note any tool behavior changes with dates so old entries stay interpretable across updates. Over a year of filings, patterns emerge revealing which destinations tighten caps silently, which document types compress reliably, and where splitting has consistently outperformed squeezing. Decisions grounded in a year of personal measurements outclass remembered advice from any guide, including this one, because they describe your files and your portals exactly.

Device notes: iPhone, Android, Windows, Mac

On iPhone and Android, scans taken with the phone camera are photos and compress well — but judge the result with a pinch-zoom to 200% before sharing, because flaws hide at fit-to-screen size on small displays.

On Windows, verify results in Edge or Acrobat at full resolution, where compression artifacts show honestly. Drag files from File Explorer straight into the tool and keep the master beside the compressed copy.

On Mac, Preview renders sharply enough to spot soft text — check footnotes there before sending. Export delivery copies from the master file each time rather than re-compressing a compressed PDF.

On any device, very large files take longer because everything runs locally in your browser: on phones, compress one file at a time and keep the tab in the foreground until the download appears.

Frequently asked questions

Why is my PDF so large?

Almost always embedded images: scanned pages saved at print resolution. Fonts and vectors are rarely the problem. Check whether your PDF is mostly scans — if so, image-based compression helps most.

Will compressed text stay selectable?

In Smart and Lossless modes, yes — Smart mode recompresses only the embedded images and leaves text and vectors untouched, so everything stays selectable and searchable. Only Maximum mode rebuilds pages as pictures, which costs selectability.

What target size should I pick for an upload form?

Read the cap stated on the form, subtract a margin, and enter that value as a custom target or pick the nearest preset under it: 100KB, 200KB, 500KB, or 1MB. The target drives Smart and Maximum modes; Lossless re-saving cannot aim at a size, and the interface states that directly.

Why did my text-only PDF not shrink at all?

Text pages store characters, fonts, and vector commands rather than pictures, so Smart mode finds no embedded images to recompress. Measured text-only fixtures return byte-identical output at 0% in 0.3 s with an under-5% notice. Remove pages or images instead of compressing harder.

Try it now — free, no signup

Smaller PDFs for email limits and faster sharing. Files stay on your device.

Open Compress PDF
How to Compress a PDF Without Losing Quality | GARGLOO