This page collects five things you do to a PDF you already have, rather than to its pages. Dark mode recolors it for reading at night. Compress makes it small enough to email. Grayscale makes it print cheaply and predictably. OCR makes a scan searchable. Repair rebuilds a file that won’t open. Every mode runs in your browser tab — the contracts, statements and records people usually run through these tools are never uploaded.
One distinction matters across all five. Dark mode, Compress and Grayscale render each page to an image and rebuild the PDF from those images, so the output looks right but its text is no longer selectable. OCR and Repair keep the original page content. Keep your original whenever you convert.
Dark mode: when dark documents actually help
Dark mode is about comfort and context, not universal superiority. In a dark room, a white PDF page is often the brightest object in view — pupils adjust to it and everything else vanishes; a dark page removes that mismatch, which is why night reading is dark mode’s strongest case. On OLED screens, dark pages also measurably extend battery life since black pixels are unlit. In bright daylight the advantage reverses: dark text on light backgrounds reads faster for most people, and astigmatism makes light-on-dark text harder for a meaningful minority.
Three themes are available: True Black, Midnight Blue and Sepia Dark. Gray and black content is recolored; colored pixels — photos, logos, colored charts — are kept and only dimmed slightly, so pictures don’t turn into negatives on digitally made PDFs.
| Source PDF | Dark-mode result |
|---|---|
| Exported from Word/Docs/LaTeX | Excellent — clean text recoloring |
| Code documentation, ebooks | Excellent — long-form reading is the core use case |
| Slides with dark themes already | Skip — double inversion helps nothing |
| Charts and data visualizations | Mixed — colors can shift meaning; check legends |
| Scans and photographed pages | Poor — the whole page is one image, so paper texture inverts too |
Compress: match the setting to where the file is going
Hitting a “file too large” wall when emailing or uploading a PDF is almost always an image problem. Phone scans are captured at far higher resolution than a screen or a printed page can show, so a five-page scanned contract can weigh 20 MB while a fifty-page text report weighs less than one. If your file is scan- or photo-heavy, there’s a lot to recover; if it’s mostly text, it’s already near its floor — and re-rendering it as images can even make it larger.
The compressor renders every page at the DPI you choose and re-encodes it as a JPEG at the quality you choose. The four presets run from 150 DPI at 85% quality down to 72 DPI at 35%.
- Email and web upload — a middle preset usually clears the common 10–25 MB attachment limit while keeping the document legible on screen.
- Print — stay at 150 DPI and high quality; lower resolutions show up as soft edges on paper.
- Portals with a size cap — compress to comfortably under the limit so a slightly larger re-export still fits.
Grayscale: where it actually pays off
On metered office printers and print-shop pricing, a color page typically costs several times a black-and-white one, and a document with a single colored logo on each page can be billed entirely at the color rate. Converting the whole file to grayscale before printing guarantees every page meters as mono, and it prints identically on every device instead of depending on each printer driver’s color conversion.
Naive color removal averages the red, green and blue channels equally, which renders yellows too dark and blues too light and can collapse adjacent chart colors into the same gray. The Rec. 601 transform used here (0.299 R + 0.587 G + 0.114 B) weights channels by perceived brightness, so dark text stays dark, highlights stay light, and a pie chart’s slices remain distinguishable. A before/after preview of page 1 shows the result before you convert; choose 200 DPI for dense, small print.
OCR: why the text layer is invisible
A searchable scan has two layers: the original page image you see, and machine-readable text positioned at the exact coordinates of each printed word. The text is drawn with zero opacity, so a pixel-perfect scan of a signed contract stays pixel-perfect — but viewers search, select and copy against the hidden layer. Replacing the image with recognized text would be destructive, because OCR is never 100% accurate; with the invisible layer, recognition errors only affect search quality, never the document.
Recognition runs on your device with Tesseract compiled to WebAssembly. The only thing downloaded is the public language model (about 15 MB, cached after the first run); nothing about your document goes the other way. For best accuracy, scan at 300 DPI, keep the page straight, and pick the document’s language — each model is language-specific. For a single photographed page rather than a PDF, the Image OCR tool does the same job; to get plain text out of an already-digital PDF, the PDF Converter’s text output is instant.
Repair: what it can and can’t do
Usually the content of a PDF that “is damaged and could not be repaired” is fine; what’s broken is its internal scaffolding. A PDF keeps a cross-reference table that tells viewers where each object lives, and an interrupted download or save can leave that index incomplete or pointing to the wrong places. Repair runs qpdf — a long-established open-source PDF library, compiled to WebAssembly — to re-scan the file for valid objects and rebuild those references, falling back to a pdf-lib rewrite if qpdf can’t process the file. A live log shows each step.
- Often recoverable — broken cross-reference tables, interrupted saves, minor structural damage where the page data survived.
- Sometimes partial — a few damaged pages may drop while the rest are recovered.
- Not recoverable — files truncated to a fraction of their size or overwritten; missing data can’t be reconstructed.
Encrypted files can’t be repaired until the password is removed, since encryption hides the structure the repair reads — use the Unlock mode of PDF Security first.
A sensible order when you need more than one
Repair first if the file won’t open. Run OCR on the original before anything that rasterizes pages, because Compress, Grayscale and Dark mode turn text into images. Compress last, once the content is final.