🔎 Textract
Every scan, every screenshot, into text. Offline. Once.
Textract at a glance
| Price | $19, once — no renewal |
| Replaces | Adobe Acrobat Pro at $19.99/mo — roughly $240 a year, every year |
| Type | Desktop app — runs offline |
| Platforms | Windows, macOS, Linux |
| Category | Finance & Admin |
Textract features
🖱️ Drag-drop OCR
Drop images (PNG/JPG/WEBP/BMP/TIFF) or PDFs, get per-page extracted text — PDFs are rasterized page-by-page automatically.
📃 Per-page results
Page thumbnails, text pane, per-page copy and copy-all, confidence indicator, .txt export.
🔍 Searchable-PDF export
Rebuilds the PDF with an invisible text layer positioned from OCR word bounding boxes — Ctrl-F-able and copy-able in any PDF reader.
📁 Batch folder mode
Point it at a folder (recursive optional), OCR everything, get a .txt per file and a summary of files/pages/words/failures.
📸 Screenshot-to-text hotkey
Global Ctrl+Shift+T opens a region-capture overlay even while minimized to tray — DPI-aware across multi-monitor setups, auto-copies to clipboard.
🌍 Multi-language OCR
English is bundled; other languages download on demand with a progress UI — the only network call the app ever makes.
🎚️ Preprocessing
Optional grayscale + adaptive threshold pass to rescue low-contrast scans.
🕘 Local history
Recent extractions persisted locally — re-copy without re-OCR, clearable any time.
$19 once — that's the whole pricing page
Adobe Acrobat Pro charges $19.99/mo — roughly $240 a year, every year. Textract is a single payment, and it keeps working whether or not we ever sell another copy.
Or get Textract inside the full suite — every app in the catalog, currently $97 for the bundle. See the bundle →
Textract guides & pages
Honest answers
Is it really free on GitHub?
Yes — the full source is MIT at github.com/bensblueprints/textract-ocr-mvp, always. $19 buys the signed installer, 1-click setup and updates instead of npm i && npm start.
How accurate is the OCR?
It's Tesseract's accuracy — excellent on clean scans, screenshots and printed documents, mediocre on handwriting or heavily skewed photos. It's not a cloud vision model, and we're not claiming it is. For sensitive paperwork the trade-off is worth it: nothing gets uploaded.
Is the searchable PDF real or a gimmick?
Real. Each page is rebuilt as the original image plus an invisible text layer positioned from OCR word bounding boxes — and the test suite round-trips the output through a PDF parser to prove the text layer extracts correctly, not just eyeballs it in a reader.