OCR Workbench

Fast, multi-language OCR for PDFs & images — with auto language detect, smart preprocessing, searchable PDF overlays, and enterprise-friendly controls.

Java • JavaFXTesseract/Tess4JNo Python requiredOffline-friendly

Core OCRLanguage HandlingPreprocessing & QualityOutput FormatsText PolishingUI & UsabilityPerformanceReporting & LogsEnterpriseDeveloperTest UtilitiesRoadmap

At a glance

• One-click OCR for PDF/PNG/JPG/TIFF/BMP

• Searchable PDF + UTF-8 text export

• Auto language detect with optional auto-install

• Smart preprocessing: rotate, deskew, binarize, DPI

• Hyphen & paragraph polish for clean text

Core OCR

• Supports multi-page PDF and scanned images (PNG/JPG/TIFF/BMP).

• Manual (user-specified codes) or Auto language modes.

• Optional Per-file subfolder output & “Open folder when done”.

• Drag-and-drop files and quick start from the Workbench.

Language Handling

• Manual entry: e.g. mar, hin, eng

• Auto-detect: probes a small crop for the best language mix.

• Language Pack Table (≈168): Installed / Update / Missing.

• In-app installer: one-click install/update of packs.

• Preflight checks & optional Auto-install before Auto mode.

Enterprise-friendly installs

• Support corporate mirror URLs & allow-lists.

• Works offline with pre-seeded tessdata.

• Datapath discovery honors -Dtessdata.dir, TESSDATA_PREFIX, common OS paths, and user profile defaults.

Preprocessing & Quality Guardrails

• Auto rotate (0/90/180/270) via projection scores.

• Auto deskew (± range/step sweep) to straighten scans.

• Auto DPI & Auto binarization for OCR readiness.

• Rescue pass on low confidence: retries rotations or language mixes.

• Confidence threshold is user-tunable.

Output Formats

• UTF-8 text (with BOM), Unicode-safe normalization.

• Searchable PDF overlay:

  • Invisible text over original page.

  • Noto fonts for Devanagari/Arabic; suitable fonts for Latin/CJK where available.

  • Optional Preserve layout for columns/tables.

Text Polishing

• Remove soft hyphens (SHY).

• Rejoin hyphenated line breaks.

• Reconstruct paragraphs using gap heuristics.

• Normalize typographic quotes & dashes.

• Preserve wide gaps as double spaces (optional).

UI & Usability

• Clean JavaFX Workbench (no FXML): file list, output folder, toggles, and logs.

• Max 5 files per job with live counter.

• Start / Stop with busy-state gating.

• Language Table button to view/install/update packs.

• Live activity log + compact toast hints.

Performance

• Per-page OCR caching (preprocessed image + languages + DPI).

• Subsampled PDF rendering for speed.

• Auto-probe on a center crop to accelerate language detection.

Reliability

Guardrails reduce re-runs: auto orientation & deskew first, rescue pass only when needed. Configurable thresholds keep throughput high with consistent quality.

Reporting & Logs

• Quality summary toast after each job (files, pages, errors).

• CSV quality report with per-page metrics (avg confidence, fallbacks, languages).

• ZIP logs export (debug output + metrics + config snapshot).

Enterprise Controls

• Block/allow network downloads per policy.

• Restrict installs to allow-listed packs.

• Mirror URL for corporate repositories.