OCR
2026-11-15
OCR turns text in images and scanned PDFs into editable text. On the desktop, recognition is handled by your vision model — no local model download; offline capability depends on whether your configured model runs locally.
Before you start
Recognition is done by an AI vision model, so configure one that accepts image input first — see AI Model Setup. Recognition requests go straight to your provider.
In PDFs
Open a scanned PDF (when the text layer is thin, the app suggests AI region recognition), box-select the area to read, and the result shows alongside the original — copyable and annotatable. PDF translation calls the vision model the same way for scanned pages: recognize first, then translate.
In images
Images in the library and clippings work the same: recognize the whole image, or drag a box around part of it — the text lands directly in the body or your materials.
Keep scans sharp and upright; for complex layouts, box regions separately. Once recognized, the text is searchable via Global Search.