PDF Text Sanitizer

A specialized tool for selectively rebuilding a PDF's searchable text layer.

This application allows you to ingest a PDF, flatten its visible text into vector outlines (using Ghostscript), and then selectively re-insert specific text as an invisible, searchable layer (OCR-style). This is ideal for cleaning up poorly OCR'ed documents, removing sensitive information while keeping the layout pixel-perfect, or selectively enabling searchability on complex documents.

Key Features

• Pixel-Perfect Outlines: Converts all visible text into vector shapes, ensuring the document looks exactly as intended without relying on installed fonts.
• Selective Searchability: Choose exactly which text remains searchable.
• Interactive UI: A keyboard-first web interface for reviewing, editing, and classifying text spans.
• Regex-Powered Classification: Batch-select text for keeping or deletion using regular expressions.
• Non-Destructive Workflow: The original PDF is never modified; a new, sanitized version is generated.

How it Works

1. Ingest: Upload a PDF.
2. Extract: The tool extracts every text span and its precise coordinates.
3. Outline: Visible text is flattened to vector outlines via Ghostscript.
4. Classify: Use regex or the manual UI to mark spans as "keep" or "delete".
5. Rebuild: Only the "keep" spans are written back into the PDF as invisible text (render mode 3) layered perfectly over the outlines.
6. Result: A pixel-identical PDF where only your chosen text is searchable and selectable.

---

Setup & Installation

1. Prerequisites (Ghostscript)

The gs command must be available in your system PATH.

• macOS (Homebrew): brew install ghostscript
• Linux (apt): sudo apt install ghostscript
• Windows (winget): winget install ArtifexSoftware.Ghostscript

2. Installation

Clone the repository and choose your preferred package manager:

git clone https://github.com/your-username/pdf-text-sanitizer.git
cd pdf-text-sanitizer

Option A: Using pip

python3 -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate
pip install -r requirements.txt

Option B: Using uv (Recommended)

uv venv
source .venv/bin/activate # On Windows: .venv\Scripts\activate
uv pip install -r requirements.txt

Running the Application

Easiest (recommended)

python start.py

Then open http://127.0.0.1:8000.

Equivalent standard entrypoint

python -m app

Optional run flags

Both commands support:

python start.py --host 0.0.0.0 --port 8000 --reload
python -m app --host 0.0.0.0 --port 8000 --reload

To disable reload:

python start.py --no-reload

Legacy uvicorn command (still works)

python -m uvicorn app.main:app --host 127.0.0.1 --port 8000 --reload

After the server is running, drag and drop a PDF into the browser to begin processing.

---

UI Keyboard Shortcuts

Key	Action
← / →	Previous / next page
g	Jump to page
Ctrl+A	Select all spans on current page
Delete	Mark selected spans as Delete
Enter	Mark selected spans as Keep
Ctrl+S	Save and download the sanitized PDF
Ctrl+Z	Undo last action
Double-click	Inline edit span text (automatically marks as Keep)

---

License

MIT