Which PDF compression level should I choose first?

Medium is a reasonable starting point. Compare the output size and inspect the smallest text. Choose low compression for more detail or high compression when a smaller visual copy matters more.

Why does low compression create a larger file?

Low describes how little compression is applied. It uses render scale 1.8 and JPEG quality 0.9, preserving more detail and generally producing a larger result than the high setting.

Can I recover searchable text after compression?

Not from this output's original text layer, because every page becomes a JPEG image. OCR can attempt to recognize the pixels later, but that is not the same as preserving the original characters.

Does this PDF compressor preserve forms and links?

No. The output pages contain flattened JPEG imagery. Interactive form controls and live hyperlinks are not recreated in the new PDF.

Will compressing always make a PDF smaller?

No. An already efficient text-based PDF can become larger after every page is replaced by JPEG image data. Compare the files before using the result.

Compress PDF Online: What Rasterizing Every Page Really Costs

Full-Page Rasterization

This tool attempts to reduce PDF size by rendering every page as a JPEG and building compressed.pdf from those images. Its three levels control render scale and JPEG quality. It works best for scanned and photo-heavy documents. The trade-off is permanent: selectable text, live links, forms, vector sharpness, and accessibility structure are not preserved. Output pages are also rebuilt as A4 portrait, and a smaller file is not guaranteed.

Open Compress PDF

When this compression method fits

Good candidates

Scanned PDFs that are already page images.
Photo-based files created from phone pictures.
Visual portfolios intended only for on-screen viewing.
Read-only reference copies when the original is stored safely.

Poor candidates

Legal, archival, or accessibility-critical documents.
Fillable forms or documents with important links.
Text PDFs that are already small and efficient.
Technical drawings or files intended for high-quality printing.

Input, output, and the three levels

Input: one readable PDF. Output: compressed.pdf with one A4 portrait output page for every source page.

Level	Render scale	JPEG quality	Trade-off
Low compression	1.8	0.9	Most detail; generally largest
Medium	1.5	0.8	Middle setting
High compression	1.0	0.6	Least detail; intended smallest

“Low compression” means less compression, not a smaller output. It renders more pixels and uses higher JPEG quality. High compression removes more image detail.

Keep the original PDF. This workflow discards document structure and cannot be reversed. An already efficient source may also produce a larger output.

How to compress and check the result

Protect the source

Keep an unchanged original. Upload a copy and start with medium compression.

Render every page

Allow time for PDF.js to draw and JPEG-encode every source page.

Compare carefully

Check file size, smallest text, landscape pages, links, forms, and print quality before using the result.

How full-page rasterization works

1. Load: PDF.js 3.4.120 reads the selected PDF from browser memory.
2. Render: each page is drawn to an HTML canvas using scale 1.8, 1.5, or 1.0.
3. Encode: canvas.toDataURL('image/jpeg', quality) converts the entire canvas to JPEG at quality 0.9, 0.8, or 0.6.
4. Rebuild: jsPDF 2.5.1 creates A4 portrait pages and stretches each JPEG to the full page width and height.
5. Download: the new Blob is saved as compressed.pdf.

This is not selective image optimization. The tool does not leave text and vectors untouched while downsampling only embedded photographs; it flattens the complete page.

File contents are processed in browser memory and are not sent to WeConvertFiles for conversion. PDF.js and jsPDF are downloaded from third-party CDNs when required.

What is permanently lost or changed

Text and accessibility

Selectable/searchable characters, document tags, reading order, alternative text, and other accessibility structure are not recreated. Assistive technology receives page images rather than structured content.

Links and forms

A hyperlink becomes visible pixels rather than a live target. Fillable fields, checkboxes, signatures, and other interactive controls are not rebuilt.

Vectors and annotations

Vector diagrams become pixels. Annotations may be flattened into the rendering when visible or omitted; they no longer retain normal interactive behavior.

Page geometry

Every output page is A4 portrait and the source rendering is stretched to fill it. Letter, landscape, square, A3, and mixed-size pages can change proportions.

Limitations and better alternatives

No target size: choose only among three presets; there is no exact MB target or direct quality slider.
No smaller-file guarantee: image pages can outweigh efficient text and vector instructions.
No selective optimization: use a dedicated optimizer when text, links, forms, and tags must survive.
No password entry: encrypted PDFs requiring a password cannot be processed through this compression interface.
Browser load: long and graphics-heavy PDFs use substantial memory. Split, process smaller parts, and merge only if the structural trade-off is acceptable.
Source-first alternative: re-export from the original application with lower image resolution whenever possible.

Privacy and data handling

File contents are processed in browser memory and are not sent to WeConvertFiles for conversion. PDF.js and jsPDF are downloaded from third-party CDNs when needed, so the workflow is not described as fully offline.

Zoho PageSense may collect consented site-usage data such as visits, clicks, scrolling, referrer, device information, and conversion events. PageSense is not given the selected PDF contents.

Troubleshooting

The result is larger than the original

The source was probably already efficient. Keep it and use a selective optimizer or re-export from the source application instead.

Small text is difficult to read

Return to the original and choose low compression, which uses a higher render scale and JPEG quality.

Search, links, or forms stopped working

That is a consequence of full-page rasterization. Use the original or another compression method that preserves document structure.

Pages changed shape

The implementation places every source rendering onto A4 portrait without preserving its original page dimensions.

A long conversion stalls

Every page is rendered and encoded in browser memory. Close heavy tabs or work with smaller document sections.

Frequently asked questions

Which level should I try first?

Medium is a useful starting point. Inspect the smallest text and compare file sizes, then use low for more detail or high for a more aggressive visual copy.

Why does low compression produce a larger file?

It applies less compression: render scale 1.8 and JPEG quality 0.9. Higher detail requires more image data.

Can I recover the original text afterward?

No. OCR could recognize some pixels later, but it would create an approximation rather than restore the source PDF's original character data and structure.

Does the number of pages change?

No. The loop creates one output page for every source page. Page content and dimensions change, but the page count remains the same.

Is this the same as PDF optimization?

No. Selective optimizers can downsample images and remove unused objects while retaining text and other structures. This implementation rasterizes the complete page.

Create a smaller visual copy

Keep your original, choose a level, and verify size, readability, and page shape before using the result.

Use Compress PDF