Supported features and limitations
These limits reflect the working browser-based PDF Tools release, not every format or capability shown on a roadmap. Keep your original files until you have checked the new result.
Inputs, outputs and performance
- PDF inputs and static JPG/PNG images are supported by the corresponding live tools. HEIC, Office conversion, OCR, password unlocking, permanent redaction, certificate signing and universal conversion are not shipped here.
- The common per-file input ceiling is 50 MiB (52,428,800 bytes) and supported PDF inputs are generally limited to 300 pages. Individual tools may impose tighter limits on total pages, number of files, outputs, resolution or memory; follow the limits displayed in each workspace.
- Large files, complex fonts, image-heavy pages and low-memory phones may process slowly, time out or fail. Device and browser capabilities vary; no universal no-crash or fidelity guarantee is made.
- JPG and PNG exports render PDF page appearance as pixels; they do not preserve selectable text or necessarily extract original embedded images. PDF-to-text exports existing text only; it does not recognize scanned text or preserve original layout.
PDF fidelity and signatures
A modified PDF copy can invalidate cryptographic signatures and can omit or change interactive forms, annotations, bookmarks, JavaScript actions and some metadata. Output verification checks supported properties; it does not prove every object or interactive feature has been preserved. The document-information inspector reports indicators rather than validating a signature certificate.
Deleting whole pages does not redact information elsewhere in a PDF. An overlaid watermark is not encryption, redaction or tamper protection. Cropping or covering content should never be assumed to securely remove it. Use a separately validated redaction workflow when secure removal is required.
Local-processing boundaries
Local browser execution does not mean offline by default: site assets come from GitHub Pages, and browser visits reveal network metadata to the host. Start over does not erase originals, downloaded files or every possible memory/cache copy. There is no account, cloud sync, persistent on-site file history, backend processing or paid conversion tier in this release.
For a practical walkthrough, read the help and FAQ. For network and temporary-memory details, read privacy and local processing.
License and reporting
The project is open source under the MIT license. Source and limitations can be inspected in the public repository. Report reproducible issues with synthetic files only through GitHub Issues; do not attach private or confidential documents to public reports.