Make a Scanned PDF Searchable (OCR Text Layer)

Invisible OCR text layer: keep the original look, gain full-text search and copy

A scanned PDF is just a stack of images - you cannot search for keywords or copy text. This tool runs an AI OCR model (PaddleOCR v5 with WebGPU acceleration) right in your browser to recognize the text on each page, then embeds the result as an invisible text layer: the page looks identical to the original scan, but becomes fully searchable and selectable.

It is a free alternative to Adobe Acrobat’s "Searchable Image" feature - and safer: contracts, IDs, and medical records never leave your device, because both recognition and writing happen locally.

English and Chinese text (including mixed) are supported, and text positions follow the OCR bounding boxes precisely, so search highlighting lands in the right place. A "replace with white background" mode is also available if you want visible, re-flowable text instead.

How to use

Open your scanned PDF
Click "OCR this page" in the preview area (page by page or in batches)
Review the recognition result and choose the invisible text layer mode
Save the PDF - the new file is searchable and copyable

FAQ

Will the invisible text layer change how pages look?

No. The text is written in invisible rendering mode, so pages look exactly like the original scan - they just gain a searchable, selectable text layer.

Where does the OCR run, and how accurate is it?

Entirely in your browser (ONNX + WebGPU/WASM); nothing is uploaded. The model is PaddleOCR v5 - very accurate on printed English and Chinese, limited on handwriting.

Is it slow?

On browsers with WebGPU (Chrome/Edge) it takes roughly 1-3 seconds per page. Without WebGPU it falls back to multi-threaded WASM, which is slower but works.

Can I process a whole book?

Yes, multi-page recognition is supported. For very large documents (hundreds of pages), process in batches to keep browser memory usage reasonable.