Technical information

Extended Tools — technical overview

Architecture, data protection, processing limits, verification and delivery for the six PDF Pro Panel 3 desktop extensions. For product information, see Extended Tools.

Updated October 10, 2026 · Coming soon, not yet on sale. Specifications may change before launch.

Requirements

Platform: PDF Pro Panel 3 desktop for Windows or macOS, version 1.15.7 or later.
Secure storage: operating-system encryption must be available. There is no plaintext fallback, so private job data is never written unencrypted.
Disk space: a 256 MB free-space reserve plus room for the job's encrypted cache, which is limited to 8 GB per job.
Network: not needed for processing. OCR language models are bundled with the app.
Processor: OCR speed depends mainly on the CPU. On a low-power 4-core Intel Celeron N5095A a full text page took roughly 12–18 seconds; faster processors take proportionally less.

Processing architecture

Isolated workers: documents are processed in separate worker processes, not in the app window. Each worker has a 768 MB working-set guard and a two-minute limit per page, and is replaced every five pages so memory use stays flat on long jobs.
One active job: only one commercial job can run at a time across all extensions; pause it to start another. Paused jobs keep their progress.
Trusted file selection: source files and export locations are chosen in native operating-system dialogs opened by the app's main process. The interface cannot supply arbitrary file paths.
Rendering: PDF pages are rendered with the native PDFium engine at 200 DPI. Images and PDF pages are limited to 16 megapixels per page; oversized pages fail explicitly rather than being silently downscaled.
OCR: Tesseract LSTM recognition (tesseract.js) with bundled English, French and Spanish models, about 5.5 MB in total, loaded locally with no model downloads.
Inputs: PDF, JPEG, PNG and multi-page TIFF. Password-protected PDFs and XFA forms are rejected; interactive AcroForms must be flattened first for image-based workflows.

Data protection

Encryption at rest: page images, recognized text and positions, review records, drafts and job checkpoints are encrypted with AES-256-GCM. The key is protected by the operating system's secure storage, and each encrypted file is bound to its own location.
Local only: document contents are never sent to a server. Licence checks transmit device and licence identifiers only.
Separate entitlements: each extension checks its own purchase on every processing and page-preview request, and keeps its own encrypted checkpoint space.
Private data removal: Clear batch, Clear all history and Remove private job data delete the encrypted cache and checkpoints. Originals and exported files are never deleted.

Integrity and recovery

Source fingerprints: every source is fingerprinted with SHA-256 and checked before and after processing and export. A changed source stops the job instead of mixing old approvals with new content.
Journaled export: each output is written to a temporary file, verified, recorded with its SHA-256 hash, then published under its final name with a hard link that cannot replace an existing file. The report is written last. A crash at any stage resumes safely; recovered files must match the recorded hash.
Crash recovery: interrupted processing or export is recovered as paused. Resuming continues from the last completed page; finished pages and saved formats are never redone.
Resource safety: storage is checked before each step. A failed checkpoint write pauses the job safely instead of continuing unrecorded.
Failure isolation: a page that fails is retried once in a fresh worker. A document that still fails is reported with its error, and the rest of the batch continues.

Review workflow

Mandatory approval: every page must be approved before export. Export of unapproved pages is refused by the processing service, not only by the interface.
Side-by-side review: each page is shown beside its details: suggestions, recognized text, fields or record data. Redaction shows the original page next to the redaction output.
Pause and resume: Pause review saves the reviewer's place. Redaction also keeps the selections and drawn areas on the open page, and Data Extraction its corrected values, as an encrypted draft. Resume review (page N) reopens exactly there, even after a restart, with no reprocessing. Approving a page clears its draft.
Exported state: exported batches are marked with the saved formats, and exporting again asks for confirmation first.

Tool specifications

Tool	Input and limits	Output	Export verification
Commercial Redaction	PDF, JPEG, PNG, TIFF. 1,024 MB per file and batch; 1,000 files; 20,000 pages per file.	Image-only PDF per document, made only from the redacted page images.	Page count, no text layer, and rendered pixels match the approved redacted images.
OCR & Document Capture	PDF, JPEG, PNG, TIFF. 1,024 MB per file and batch.	Searchable PDF, Word (.docx), Excel (.xlsx), Rich Text (.rtf) or EPUB 3 e-book (.epub).	PDF: page count, visible pixels, and a text layer containing every recognized character (its order follows the page layout). EPUB: a reflowable book with one section per page and a table of contents from detected headings. Word, Excel and Rich Text: generated text, page boundaries and illustrations are recorded; layout and tables need review in the destination application.
Legal Document Production	PDF only. 100 MB and 2,000 pages per PDF; 1,024 MB per batch.	Numbered PDFs with a 22-point footer outside the original page area, plus a CSV production index.	Page count, and each page's Bates label read back from the output.
Forms Automation	One AcroForm template (25 MB, 100 pages, 1,000 fields) and a CSV (2 MB, 1,000 records, 100 columns).	One flattened PDF per valid record, plus an exception index.	Fields flattened, approved values present in the output text, required fields filled.
Data Extraction	PDF and images. 1,024 MB per batch; up to 30 labelled fields.	CSV, one row per document, with formula-like values neutralized.	Operator-approved values only.
Document Sanitization	PDF and images. 1,024 MB per file and batch.	Fresh image-only PDFs with no imported PDF objects.	Page count, matching visible pixels and no searchable text.

Redaction detection

Industries: one or more of 50 draft industry profiles. A batch with several selected suggests the combined set of entity types of all of them; the first selected is recorded as the primary profile. Profiles are starting points, not compliance certifications.
Regions: one or more of Canada (Social Insurance Number), United States (Social Security Number), United Kingdom (National Insurance and NHS numbers, UK phone numbers), Ireland (PPS number), Australia (Tax File Number, Medicare number, Australian phone numbers), France (NIR social security number, French phone numbers), Spain (DNI and NIE) and Mexico (CURP). Every identifier is matched by its format and then confirmed by its official check digit or structure rule, so ordinary numbers of the same shape are not suggested. A validated identifier keeps its own label even when it also looks like a phone number.
Other suggestions: email addresses, payment card numbers (Luhn-checked), North American phone numbers and international phone numbers written with a leading + (8 to 15 digits), plus labelled names, addresses, birth dates and record numbers in English, French and Spanish.
Compatibility: batches created before multiple selection keep their single industry and region (Canada and US) and continue to scan the same way after an update.
Reports: each export report records the industries, regions and document languages used.

OCR Auto Batch

Intake: a multi-page file, several files, or a folder (optionally including subfolders up to five levels deep). Up to 1,000 supported files per batch. Hidden files, shortcuts, empty and unsupported files are skipped and counted. Pages are ordered by natural name sort.
Page clean-up (on by default): each photo is first checked for a two-page spread, which is split when both halves carry text and a wide white gutter or a dark spine shadow lies near the middle (two-column pages have narrower gaps and are not split). The layout is decided once per document and saved, so resuming makes the same choices. Each page is then trimmed of dark edges, straightened by the angle at which its text lines align best (up to 6°), and brightened if the photo is dim (never darkened).
Processing: unattended OCR with progress and time remaining. Pages with confidence below 70% or no recognized text are flagged.
Review: side-by-side approval by default, with pause and resume. Review can be turned off for unattended runs.
Compiled output: by default every approved page of every input is written, in order, into one document per selected format. Word is the default; Searchable PDF, EPUB e-book, Rich Text and Excel are optional. The location is asked for once, with Save As, and existing names are refused rather than replaced. One document per file is also available.
Export again: a finished batch can be saved in another format without reprocessing or re-review.

Outputs and reports

Verification report: each output has a .report.json with input and output SHA-256 fingerprints, settings, language, time, page count where applicable, and the scope of the checks performed. Reports do not include extracted text or form values.
Batch index: per-document batches produce a CSV index with outputs, status, errors and fingerprints, plus Bates ranges where applicable. Spreadsheet formula characters are neutralized in all CSV output.
Naming: per-document outputs use internal identifiers, so source file names are not revealed. Compiled Auto Batch documents use the name the user chooses.
Viewing: saved outputs up to 128 MB open in the app's PDF viewer, Word editor or spreadsheet editor.

Licensing and delivery

Separate entitlements: each extension has its own entitlement and is purchased individually. Purchases never unlock the base app or other extensions.
Trial and purchase options: a 3-day free trial per extension, and a choice of one-time purchase or yearly subscription.
Direct desktop editions: extensions are delivered as signed .ppmodule packages. Before installation the app verifies the release signature (RSA-SHA256, pinned public keys), version, compatibility, package size and SHA-256 hash, module identity and entitlement. Ownership is embedded in the existing device-bound, signed offline licence certificate.
Store editions: extensions are included in the app package and unlocked by native Microsoft Store or Mac App Store purchases.
See the add-ons page (licence, price, trial and FAQ) and the extension purchase terms.

Known limitations

OCR languages: English, French and Spanish. Searchable PDF text uses Latin (WinAnsi) characters; an unsupported recognized character fails the export explicitly rather than being substituted.
Word, Rich Text and Excel outputs reconstruct content. Identical fonts, tables and pagination are not guaranteed; complex tables and low-contrast artwork may need manual correction.
Image-based outputs (Redaction, Sanitization) do not retain searchable text, accessibility tags, interactive fields or digital signatures.
EPUB output reflows the recognized text; the printed page layout, fonts and tables are not reproduced.
Not currently included: watched (hot) folders, barcode separation, saved company templates and mappings, table extraction, parallel OCR on several processor cores, and OCR languages other than English, French and Spanish.
Workflow screens are currently in English.