DocFlow

Convert PDF, Images & Office Documents to DOCX with OCR — Local, Private, No Upload

100% Local · Privacy-First · No Document Uploads

English · 简体中文

---

DocFlow is a zero-code, fully local document conversion tool that batch-converts PDF, images, and Office documents into editable DOCX files. It features OCR text recognition (RapidOCR), AI image description (Qwen-VL), and table restoration — all running locally with no document ever leaving your machine.

✨ Features

Feature	Description
Multi-format Input	PDF, DOCX, XLS, XLSX, JPG, PNG, GIF, BMP, WebP, TIFF
Batch Processing	Drag & drop multiple files, track progress per file
OCR Text Recognition	Powered by RapidOCR — extracts text from images, scans, and PPT screenshots; supports Chinese-English mixed text
AI Image Description	Enhanced mode uses Qwen-VL local multimodal model to generate semantic descriptions for images
Table Restoration	Converts image-based tables into editable DOCX tables
Bilingual UI	Chinese / English one-click switch
100% Local & Private	No internet required, no document uploads — your data stays on your machine

🆚 Why DocFlow?

	DocFlow	Online Converters	Adobe Acrobat
Privacy	✅ 100% local, no upload	❌ Files uploaded to server	⚠️ Partial
OCR	✅ Built-in (RapidOCR)	⚠️ Some have it	✅ Yes
AI Image Description	✅ Qwen-VL	❌ No	❌ No
Batch Conversion	✅ Free	⚠️ Often paid	✅ Paid
Cost	✅ Free & Open Source	⚠️ Freemium	❌ Expensive
No Account Needed	✅	❌ Sign-up required	❌ Subscription

🚀 Quick Start

Prerequisites

• OS: Windows 10 or later
• Python: 3.10 – 3.13 (Download, check "Add Python to PATH")
• Enhanced mode (optional): NVIDIA GPU with 10GB+ VRAM

Installation

Option A — One-click install (recommended):

Double-click install.bat and wait for dependencies to finish.

Option B — Manual install:

# Create virtual environment (recommended)
python -m venv .venv
.venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

Run

Double-click run.bat (or run run.ps1 in PowerShell). Your browser will open at http://localhost:7860 automatically.

📖 Usage

Two Conversion Modes

Mode	Features	Hardware
Standard Mode	Preserve original images + OCR text recognition	Any computer
Enhanced Mode	Adds AI image description (Qwen-VL)	NVIDIA GPU, 10GB+ VRAM

Steps

1. Add files — Click or drag files into the upload area
2. Select mode — Standard (any PC) or Enhanced (requires NVIDIA GPU)
3. Set output path — Click "Browse" to choose output folder
4. Start conversion — Click "Convert" and watch real-time progress
5. View results — Click "Open" to jump to the output folder

🏗️ Project Structure

DocFlow/
├── app.py                  # Flask backend main application
├── run.bat                 # Windows launcher
├── run.ps1                 # PowerShell launcher
├── install.bat             # One-click installer
├── requirements.txt        # Python dependencies
├── LICENSE                 # MIT License
├── README.md               # English docs
├── README.zh-CN.md         # Chinese docs
├── assets/                 # Icons & images
├── core/                   # Core conversion modules
│   ├── converter.py        # Main converter (batch & single)
│   ├── parser.py           # Document parsing (PDF/DOCX/XLSX)
│   ├── ocr_engine.py       # OCR engine (RapidOCR)
│   ├── image_describer.py  # AI image description (Qwen-VL)
│   └── docx_builder.py     # DOCX generator
├── utils/                  # Utilities
│   ├── helpers.py          # Helper functions
│   └── i18n.py             # Internationalization
└── web/
    └── index.html          # Frontend UI

🔧 Tech Stack

Component	Technology
Backend	Flask 3.x
Frontend	Vanilla HTML / CSS / JS
PDF Parsing	PyMuPDF (fitz)
OCR	RapidOCR (ONNX Runtime)
AI Description	Qwen-VL-Chat-Int4 (via ModelScope)
DOCX Generation	python-docx
XLSX Parsing	openpyxl

❓ FAQ

Q: Why can't .doc files be converted?
A: python-docx does not support the legacy .doc format. Please save as .docx in Word first.

Q: Enhanced mode says unavailable — what should I do?
A: Enhanced mode requires an NVIDIA GPU with 10GB+ VRAM. Switch to Standard mode, which works on any computer.

Q: Where are the converted files?
A: By default in the output/ folder. You can change the output path in the UI and click "Open" to access it directly.

Q: Does it support macOS / Linux?
A: Currently Windows only. The core code is cross-platform, but the launcher scripts (.bat/.ps1) and system calls need adaptation. PRs welcome!

Q: Is it really 100% offline?
A: Yes. Standard mode needs no internet at all. Enhanced mode downloads the Qwen-VL model once (on first run), after which it runs fully offline.

🤝 Contributing

Contributions are welcome! Feel free to:

• 🐛 Report bugs or suggest features
• 🔀 Submit a Pull Request
• ⭐ Star the project if you find it useful
• 📢 Share it with others who might benefit

📄 License

MIT License — free for personal and commercial use.

---

If DocFlow helps you, consider giving it a ⭐ — it helps others discover it too!