Searchable
Open the app: Make a scan searchable → (products.documentize.app)
In one sentence
Searchable takes a scanned PDF (pages that are just pictures) and gives you a PDF that looks identical but has an invisible text layer on top, so you can search, select and copy the words.
What makes it unique
- It keeps the page picture and adds the text underneath, so the document looks the same as the scan. OCR Image returns only text, and Extract Text cannot read pictures of text at all.
- It has scan-repair options built in: automatic straightening (deskew), background and watermark cleanup, a "source photo is sideways" switch, and an optional AI correction pass for recognition mistakes.
- It can also output editable Word, Excel or PowerPoint files built from the recognized text, and for Excel it can extract tables only.
When to use it / when not to
- Use it when you have a scanned contract, book chapter, invoice or archive and you want to find words in it, copy passages, or feed it to other tools (Search, Ask PDF, Extract Text).
- Do not use it on a PDF that already has real text; it would redo work and may change nothing useful. Test first: if you can select words in the PDF, it is already searchable.
- Do not use it if you just need a .txt from a photo: OCR Image is simpler.
- Not for locked files. Remove the password first with Unlock.
Before you start
- The web page takes PDF files. If you have photos, first turn them into a PDF with Merger.
- Know the language of the text.
- Fix upside-down or sideways pages first (Rotation); orientation errors hurt recognition more than anything else.
- Very long documents are refused with a message asking you to split the file into smaller parts (the page limit is set by the plan, so check the site). Use Splitter.
Step by step
- Open Searchable and drop your scanned PDF (or several).
- Choose the document language.
- Open Show all settings: leave Deskew on for photographed pages, tick Source photo is sideways if the pages are rotated 90 degrees, and tick Remove background/watermark for shaded paper or stamps.
- Pick the Output format (Searchable PDF by default).
- Press Transform, then download the result. Transform again starts over.
📷 Screenshot placeholder: Searchable with all settings open
Options explained
| Option | What it does | Default | Recommended |
|---|---|---|---|
| Document language | Which language the recognition expects | English | Always match the document |
| Deskew image | Straightens each page slightly before recognition | on | Keep on, especially for phone photos |
| Source photo is sideways (rotate 90°) | Turns pages that were scanned or photographed on their side | off | Use only if you know the pages are sideways; there is no automatic orientation detection |
| Remove background/watermark | Cleans shadows, paper colour and watermark bleed-through before recognition | off | On for yellowed, shaded or stamped pages |
| Correct OCR errors with AI | A language model proofreads the recognized text line by line before it is embedded | off | Try it when plain recognition makes many errors; always spot-check |
| Landscape output page | Output pages are made landscape | off | Rarely needed |
| Output format | Searchable PDF, Editable Word (.docx), Editable Excel (.xlsx) or Editable PowerPoint (.pptx) | Searchable PDF | PDF to keep the look; Office formats to edit the content |
| Extract tables only | With Excel output, puts only detected tables in the workbook, skipping full-page text | off | On for scanned tables and reports |
| Background cleanup (advanced) | Three numbers (light-background ratio 0-1, saturation threshold 0-255, near-white level 0-255) that tune the cleanup | built-in defaults | Leave alone unless cleanup is wrongly treating pages as text or photo |
What you get
- A PDF with the same pages plus a hidden text layer (or a .docx/.xlsx/.pptx if you chose one). Several input files come back in a ZIP.
- Your original is never modified.
- In the Office formats the page picture is dropped and the text is flowed into the document, so it is editable but looks different from the scan.
- If adding the text layer fails midway, the service falls back to delivering the cleaned pages without the layer rather than nothing.
Quality & limits
- Recognition is an estimate: scan quality, language choice and fonts decide accuracy. Poor input gives wrong words that look right in the file.
- The AI correction is generated and can change words; do not use it where every character must be exact (names, numbers, legal wording) without checking.
- The file can get larger because the text layer is added; run Compressor afterwards if needed.
- Free-plan time limits can stop a batch early and leave a "continuation" to finish the remaining files later (about 12 hours of cool-down by default). Results are kept for about two days.
- Complex layouts (multiple columns, tables) are only approximately preserved in Word/Excel/PowerPoint output.
Works well with
- Use before: Rotation, Deskew (a standalone straightener), Unwatermark, Unlock, Merger (photos to PDF).
- Use after: Search, Extract Text, Add TOC, Ask PDF, Compressor.
- Recipes: Photos to searchable PDF, Digitise an archive.
Common mistakes & fixes
- "I ran it and the PDF looks the same" → that is correct; the text layer is invisible. Press Ctrl+F in your viewer to check.
- "Words are wrong or missing" → wrong language, sideways pages, or low-resolution scan. Fix those and rerun; try background cleanup and AI correction last.
- "It says to split the file" → the document is too long for one pass; split it, process the parts, merge them back.
- "Text is selectable but the order is odd" → columns and tables are read approximately; use Parser or Excel output for structure.
Same thing elsewhere
- CLI plugin: Documentize.OcrSearchable (needs Tesseract installed locally).
- MCP tool: MakeSearchable.
- See Beyond the browser.
Try it
Scan or photograph two pages of a printed document and save them as a PDF. Run Searchable in English with defaults, open the result and press Ctrl+F for a word you can see on the page. Repeat with Editable Word output and compare the two files: one looks like the scan, the other is something you can retype into.