ToolwebHow to use this tool
- Choose one PDF and check its original size and page count.
- Choose a mode. For either JPEG mode, review resolution, quality and the loss notice.
- Inspect processed size and pages. Check whether the default download is the original or processed file before saving.
A PDF mixing searchable text and scanned pages
- Input and settings
- Two text pages plus one photo page without text
- Result
- Selective mode copies the two text pages and rebuilds the photo page as JPEG at the selected DPI and quality
Classification uses PDF.js text extraction. Size reduction is not guaranteed; if processing grows the file, the original remains the default.
How the three modes differ
Structure rebuilding saves with object streams without redrawing pages. It does not resample photos or convert existing text into images. An already optimized PDF may not shrink.
Keep-text mode copies pages where PDF.js finds non-empty text, and rebuilds only the other pages as JPEG. Use it to retain searchable pages while reducing text-free photo pages. Scans containing an OCR layer are copied, so their images are not recompressed.
All-JPEG mode rasterizes every page. It can help photo-heavy files, but removes searchable/selectable text. Both JPEG modes lose links, annotations, forms, vectors and accessibility information on rasterized pages, and require acknowledgment.
Selective compression is not scan recognition
This tool does not use AI to identify scans. A page without extracted text becomes JPEG even if it contains tables, vector drawings or outlined lettering. No OCR or new text recognition is performed. Zoom into small text and thin lines in the downloaded processed PDF.
Keep the original when processing grows the file
Actual byte counts, difference and percentage are compared. A smaller processed file becomes the default download; an equal or larger result leaves the byte-identical original as the default. A larger processed PDF remains available through a comparison link. Downloading the original keeps its original text and document structure.
Previews show only the first page of the original and processed files. The right preview is always processed, even when the default download is original. Document-wide forms, signatures, bookmarks and accessibility structures are not guaranteed in a rebuilt PDF; retain important originals.
Quality and resource limits
72 DPI is low-resolution screen output; 108 and 144 DPI retain more pixels but may increase size. Limits are 30 MB input, 30 pages, eight million raster pixels per page, forty million overall, 75 MB output and 90 seconds overall. Lower resolution or use fewer pages when limits are exceeded.
Common questions
Can text still be searched after selective compression?
Pages where PDF.js extracts text are copied without redrawing, retaining existing searchable text. Other pages become images without new OCR. All-JPEG mode removes searchable text on every page.
Which file should I keep if processing made it larger?
The default download automatically switches to the original. Check both byte counts and optionally download the processed comparison file.