Why local processing
Upload-based PDF sites send the file to a server so that server can rewrite it. That is convenient, and it also means the operator, their processors, and anyone who obtains those logs can see the document. ZANCTA’s implemented PDF and image tools read the file you selected in the browser and write the result in the browser. The bytes for those workflows are not posted to ZANCTA for processing.
What still leaves the device
The application itself is downloaded like any website: HTML, JavaScript, fonts, and images. Analytics, if you allow it, may record that a tool finished — with the tool name only. Accounts, email, and payments (when live) send account and billing data to those providers. None of that is a substitute for uploading the PDF. If you want to check a browser tool yourself, use the browser file-upload verification guide.
OCR
Image OCR runs recognition in a Worker. English language data is bundled. Hindi, Bengali, Tamil, Spanish, French, and German packs load only for Premium, and only when selected. Scanned PDF OCR is also Premium and capped at 20 pages. Handwriting, dense tables, tiny type, and mixed scripts will fail or produce garbage. The recognized text is not sent to an OCR API. Accuracy is not human-level.
Start at Image OCR or the browser OCR guide.
PDF text extraction
PDF Text Extractor copies text that is already embedded in a text-native PDF. A scanned “PDF” that is only images has no such text; the tool reports that instead of inventing words. For scans, use Image OCR Local OCR Power (Premium, up to 20 pages) rather than expecting this extractor to invent text. The PDF text extraction vs OCR guide shows how to choose between the two.
Start at PDF Text Extractor.
Images and metadata
Compression, conversion, and resize stay in the browser for JPG, PNG, and WebP. HEIC and SVG are out of scope. EXIF cleaning re-encodes supported images so typical GPS and camera tags do not survive; it cannot prove a downstream site will not re-tag a file you later upload elsewhere. See Remove EXIF before sharing.
PDF size and pages
Compress PDF without uploading explains the object-stream rewrite — images inside the PDF are not transcoded. Split PDF without uploading extracts a range into one new PDF. Merge PDFs without uploading combines files in the same local model.