Making a PDF smaller is a trade-off among file size, image detail, and readable or selectable text. We tested five online compressors with three fictional PDFs, measured every file we could download, and noted when a result was visible only on a web page. If you need a particular size, 200 KB and 50 KB are examples of targets you can choose—not one-size-fits-all presets.
Disclosure: RedPanda Compress publishes this blog and is one of the tools tested. We report its downloaded file sizes alongside the others and distinguish a smaller file from better image quality at the same size.
Start with the file, then choose a size
- For a scanned page: reducing the embedded image made the biggest difference. Smallpdf Basic returned 145,048 bytes at roughly 130 ppi, RedPanda returned 153,522 bytes at roughly 165 ppi with a user-entered 200 KB limit, and iLovePDF returned 166,640 bytes at roughly 150 ppi. All remained legible in our 150-dpi render; the smallest file did not retain the most pixels.
- For a report with a large image: RedPanda returned 80,837 bytes when we entered 200 KB and 39,395 bytes when we entered 50 KB. iLovePDF’s default setting returned 42,932 bytes. All three kept two pages and selectable text, but their image settings differ. The 200 KB run kept more photo texture; the 50 KB run shows RedPanda can also make this report smaller when the size goal calls for it. These byte counts are not an equal-quality ranking.
- For a text-and-vector-heavy PDF: image quality controls have little to work on because the file contains no embedded photos. RedPanda losslessly repacked supported page streams, reducing it from 553,611 to 426,233 bytes; iLovePDF returned 422,331 bytes. Both kept 20 pages and selectable text. Neither reached the example 200 KB target.
- For private documents: RedPanda processes the PDF in the browser; the other services in this comparison upload files to their servers. Decide whether a contract, ID scan, or medical record belongs on an upload service before selecting a tool.
How we tested
We generated three PDFs with no real personal information: a one-page A4 scan at 300 dpi (4,506,788 bytes), a two-page report with selectable text and one large photo-like image (984,175 bytes), and a 20-page appendix of text and vector marks without embedded photos (553,611 bytes). We retained the inputs and generator. They exercise different ways a PDF takes space: raster pixels, a mixed page, and drawing instructions.
On October 21, we entered a 200 KB size limit in RedPanda for all three PDFs, then separately entered 50 KB for the mixed report. These are user-chosen limits, not RedPanda presets. On October 20, we tried PDF24 at its displayed 144-dpi/75% defaults, iLovePDF at its default setting, Smallpdf on free Basic, and Adobe on High. We made one run per available setting; the rows below therefore compare different settings and image quality, not compression efficiency at matched quality. For downloaded files, we measured actual bytes, checked PDF validity and page count, rendered pages at 150 dpi, and extracted text where the source had selectable text. PDF24 and Adobe outputs could not be downloaded, so their interface figures are not verified file sizes. This is a small test of these files and settings, not a universal ranking.
Open the exact test PDFs
These are the original, synthetic input files used for the measurements below—not compressed results. They contain no real personal information. You can inspect them, rerun the compression test, and verify each download against the published checksums.
- Scanned checklist (PDF) — 1 page, 4,506,788 bytes; an image-only A4 page scanned at 300 dpi.
- Image-and-text report (PDF) — 2 pages, 984,175 bytes; selectable text and one photo-like image.
- Text-and-vector appendix (PDF) — 20 pages, 553,611 bytes; selectable text and vector marks, with no embedded photos.
For reproducibility: SHA-256 checksums and the Python fixture generator.
Measured sizes and the detail behind them
Different settings, not a quality ranking. Each table reports one source PDF. Results marked “B” are measured downloads; “displayed” figures could not be checked. The image-detail note belongs with the size—read both before deciding which output suits your document.
Scanned page — 4,506,788-byte source
| Tool and actual setting | Output size and scan detail |
|---|---|
| RedPanda Compress · user-entered 200 KB goal | 153,522 B downloaded · about 165 ppi |
| Smallpdf · free Basic | 145,048 B downloaded · about 130 ppi, the fewest pixels here |
| iLovePDF · default setting | 166,640 B downloaded · about 150 ppi |
| PDF24 · 144 dpi / 75% | 227.83 KB displayed† · file and image quality unverified |
| Adobe Acrobat online · High | 59.44 KB displayed‡ · file and image quality unverified |
Image-heavy report — 984,175-byte source
| Tool and actual setting | Output size and image detail |
|---|---|
| RedPanda Compress · user-entered 200 KB goal | 80,837 B downloaded · 1,493 × 954 px image, more photo texture |
| RedPanda Compress · separate user-entered 50 KB goal | 39,395 B downloaded · 1,353 × 864 px image, stronger JPEG compression |
| iLovePDF · default setting | 42,932 B downloaded · 1,053 × 708 px image, smoother photo |
| PDF24 · 144 dpi / 75% | 71.81 KB displayed† · file and image quality unverified |
| Adobe Acrobat online · High | 9.33 KB displayed‡ · file and image quality unverified |
| Smallpdf · free Basic | Not run after the observed free-use gate* |
Text and vector appendix — 553,611-byte source
| Tool and actual setting | Output size and verification |
|---|---|
| RedPanda Compress · user-entered 200 KB goal | 426,233 B downloaded · 20 pages and selectable text; goal unmet |
| iLovePDF · default setting | 422,331 B downloaded · 20 pages and selectable text |
| PDF24 · 144 dpi / 75% | 428.53 KB displayed† · file unverified |
| Adobe Acrobat online · High | Not run after the observed sign-in gate‡ |
| Smallpdf · free Basic | Not run after the observed free-use gate* |
Read the setting and detail with each size: RedPanda’s 200 KB and 50 KB report results are separate choices; a smaller byte count alone does not mean better image quality. Approximate scan ppi and report image dimensions describe retained pixels, not every aspect of visual quality. Only byte counts marked “B” came from downloaded files. *After the Smallpdf scan download, the page showed a Pro upgrade or an approximately 12-hour wait before another operation; we cannot infer every visitor’s allowance. †PDF24’s figures came from its results page; Chrome blocked its results frame before we could download and check the files. ‡Adobe displayed these figures, but the scan download failed, the report download required sign-in, and the next upload was gated. We could not verify either Adobe PDF’s actual bytes, pages, text, or image quality. “KB” labels may be rounded or use different definitions: check the saved file if your destination enforces a limit.
What the sizes hide
A scan is one big image. The three downloaded scan outputs all fit under 200,000 bytes and retained every visible row, checkbox, and fine-print line in our 150-dpi rendering. They got there by reducing the embedded image from the source’s 300 pixels per inch: roughly 165 ppi for RedPanda, 150 for iLovePDF, and 130 for Smallpdf. That is adequate for this test page, but a faint stamp, handwriting, or small seal could fare differently. Zoom in before uploading a compressed official document. The source was image-only, so its words were not selectable before compression and did not become selectable afterward; compression is not OCR.
The image settings explain the size difference. In the mixed report, iLovePDF reduced the shared image to 1,053 × 708 pixels and returned 42,932 bytes. RedPanda’s 200 KB run kept it at 1,493 × 954 pixels, with more visible texture, and returned 80,837 bytes. Their encoded image streams were 40,216 and 78,488 bytes respectively, accounting for essentially the entire whole-file gap. A second RedPanda run, this time with a user-entered 50 KB target, returned 39,395 bytes and kept a 1,353 × 864-pixel image with stronger JPEG compression. All three files retained two pages and selectable text. The 50 KB result proves the 80,837-byte result was a quality-setting choice rather than a floor; it does not prove equal visual quality. Compare the rendered pages at the size your recipient will use.
Non-image bytes can shrink too. The text/vector appendix has no embedded photos, so changing image resolution or JPEG quality cannot help it. RedPanda’s released version losslessly repacks its ASCII85-plus-Flate page streams, reducing the PDF from 553,611 to 426,233 bytes while keeping the decoded drawing instructions identical. iLovePDF returned 422,331 bytes, just 3,902 bytes smaller, and also kept all 20 pages and selectable text. Neither output met the example 200 KB limit. This is not proof that the file cannot shrink further: another local PDF optimizer produced 392,001 bytes in a separate test. For a much smaller goal, remove unneeded pages or content, re-export the source, or ask the recipient for a higher limit.
Which tool should you use?
RedPanda Compress is a direct choice when you want to enter your own size goal and keep processing on your device. We tested both 200 KB and 50 KB as examples, not fixed presets. The report went from 80,837 to 39,395 bytes as we asked for a smaller file; inspect the image before choosing the smaller version. Its current version also losslessly repacks supported non-image streams, reducing our text-heavy file substantially, though it could not reach 200 KB. Processing locally can use noticeable CPU for larger PDFs.
Smallpdf Basic gave the smallest verified scan output, 145,048 bytes, despite estimating about 486 kB first. It looked readable, but a free-use gate appeared after that download. Moderate and Strong are Pro features. Check the saved file rather than the estimate.
iLovePDF (default setting) accepted all three PDFs in one batch without a login and returned a ZIP. For the mixed report it chose a 1,053 × 708-pixel image and a 42,932-byte file. That is smaller than RedPanda’s 200 KB-target output, which kept more photo detail; it is larger than RedPanda’s separate 50 KB-target output. These are different image settings, so compare legibility and texture at the file size you actually need. iLovePDF’s presets are convenient, but there is no box for a specific size such as 200 KB or 50 KB; download and check each result. Its text-heavy output was 3,902 bytes smaller than RedPanda’s.
PDF24 exposes dpi and JPEG quality controls for free. At its 144-dpi/75% defaults, it displayed 227.83 KB for the scan and 71.81 KB for the report. You can adjust those controls toward a smaller file, but Chrome blocked its results frame before download, so we cannot confirm actual bytes or image quality.
Adobe Acrobat online offers High, Medium, and Low compression, without a typed size target. High displayed 59.44 KB for the scan and 9.33 KB for the report, but the scan download failed and the report download required sign-in. The next upload was gated too. With no files to inspect, neither displayed figure is a verified pass.
A practical workflow for any size goal
- Choose the goal that matters to your recipient or storage budget. A portal might require less than 200 KB or 50 KB; those are different targets, not universal “best” settings. Leave room below a stated cap.
- Keep the original. For a document with a digital signature, any PDF rewrite can invalidate that signature; confirm what the recipient accepts before compressing.
- Use a size you type or adjust a quality preset, then inspect the downloaded file size. If the destination compares actual bytes, remember that some tools use 1,024-byte units when they display KB. A progress estimate is not the answer.
- Open every page at normal and zoomed reading size. Check small print, stamps, handwriting, charts, and page count. If the document was born digital, also try selecting a sentence of text.
- If the file cannot fit, reduce the number of pages or re-export the source rather than making important details illegible. Our 200 KB PDF guide walks through that decision.
FAQ
How should I choose between a 200 KB, 50 KB, or other target?
Use the recipient’s limit or the amount of space you can afford, then choose the largest file that preserves the detail you need. A strict 200 KB portal may compare against 200,000 bytes, and a 50 KB portal against 50,000 bytes; some tools display KB in 1,024-byte units. Leave headroom, inspect the downloaded file’s actual byte count, and compare rendered pages before uploading.
Why did a 20-page PDF fail even though it was mostly text?
Our dense vector appendix had no embedded image to shrink. RedPanda’s lossless stream repacking cut it from 553,611 to 426,233 bytes, and iLovePDF returned 422,331 bytes. Both remained above the 200 KB example goal; changing image quality alone cannot solve this fixture.
Will compressed text remain selectable?
It did in our born-digital report and vector appendix outputs we could download. The scan was just pixels before compression and remained pixels afterward. Run OCR separately if the recipient needs searchable text.
Is uploading a sensitive PDF safe?
That depends on the document and your organization’s rules. PDF24, Smallpdf, iLovePDF, and Adobe’s online service process uploads on their servers; read their current retention and privacy policies before sending sensitive material. RedPanda Compress runs PDF compression inside your browser, so the document stays on your device. The tradeoff is local CPU use, and no tool can guarantee that every PDF will meet your chosen size without losing necessary detail.
Fei is a skilled software engineer. He previously worked at Google and now at a startup. His expertise includes web media processing, cloud architecture, complex algorithms, and AI training and deployment. Beyond work, Fei enjoys diving into new knowledge and is a big fan of strategy games.
Leave a Reply