All guides

Files & browser processing

What browser image and PDF tools can preserve

Keeping a file on your device reduces upload exposure, but every transformation has a preservation boundary. Know whether a tool copies pages, redraws pixels, or rebuilds document structure.

Published and reviewed by Handiwork11 min readReviewed July 22, 2026

Key takeaways

  • Local processing describes where work runs, not which file features survive.
  • Canvas resizing creates new encoded pixels and commonly drops source metadata.
  • PDF page copying can preserve visible pages while changing document-level features.
  • Keep originals and validate the downloaded result with the workflow that will consume it.

Local processing and preservation are separate claims

The File API allows a page to read a file selected by the user. JavaScript can then transform bytes, draw decoded images, or construct a new PDF without uploading the working file. That architecture answers where the operation runs; it does not promise that every source feature appears in the result.

A transformation should name what it reads, what it changes, and what it writes. Users can then decide whether a local convenience tool fits a casual task or whether an approved desktop, archival, legal, or prepress workflow is required.

Canvas resizing creates a new image

An image resizer normally decodes a source, draws pixels to a canvas at new dimensions, and asks the browser to encode a new Blob. Downscaling discards detail; upscaling interpolates pixels without restoring information. Free resizing can change proportions, while aspect-ratio fitting may produce dimensions smaller than one side of the requested box.

EXIF camera data, XMP, ICC profiles, animation, and format-specific features are not automatically copied by a simple canvas workflow. Browser encoders and maximum canvas sizes also differ. Inspect the actual downloaded MIME type, color, transparency, sharpness, dimensions, and metadata.

PDF pages are not the whole document

A PDF can include pages, fonts, images, form fields, annotations, bookmarks, page labels, attachments, layers, metadata, scripts, accessibility structure, and signatures. A page-copying library may produce the visible sequence a user expects while not transferring every document-level relationship.

Merging or extracting pages creates a new document. A cryptographic signature on an input does not certify that new output. Page numbers used by software can also differ from printed labels such as Roman-numeral front matter.

Structural optimization is not image compression

Re-saving a PDF with object streams can rewrite cross-reference and object structures. It may reduce some overhead, but a scanned PDF dominated by images will not become substantially smaller unless those images are downsampled or recompressed.

A trustworthy tool should not guarantee compression from a structural re-save. Compare the byte size before and after, but also inspect visual quality, links, forms, attachments, metadata, and signatures. A smaller file that loses a required feature is not a successful result.

Rasterization and metadata cleaning are narrow operations

Rendering PDF pages to PNG turns text and vector graphics into fixed-resolution pixels. Searchability, selectable text, links, forms, and accessibility structure do not survive in the images. A ZIP of PNG pages is a visual export, not an equivalent PDF replacement.

Likewise, clearing common Title, Author, Subject, Creator, and Producer fields is not full anonymization. Other XMP, comments, form values, attachments, hidden content, signatures, filenames, or visible text can identify a person or organization.

A file-output verification checklist

Keep the original, use a non-sensitive sample first, and record the expected page count, dimensions, file type, and required features. After processing, reopen the result in the application that will actually consume it and compare representative pages at normal and high zoom.

For important documents, test search, links, forms, bookmarks, attachments, print output, metadata, accessibility, and signatures as applicable. Browser memory failures are also possible with large inputs because source bytes, decoded content, canvases, and the output can coexist in memory.

Primary sources

These standards and public-agency references support the factual guidance above. Links open at the original publisher.