---
title: Document OCR
emoji: 📄
colorFrom: indigo
colorTo: blue
sdk: gradio
sdk_version: 6.15.1
app_file: app.py
python_version: "3.12"
short_description: Persian and English OCR for images and PDFs
license: apache-2.0
pinned: false
startup_duration_timeout: 1h
models:
- PaddlePaddle/PaddleOCR-VL-1.6
tags:
- ocr
- document
- persian
- farsi
- english
- pdf
- table
- markdown
---

# Document OCR

Public document OCR with an English interface and a full Persian UI. Persian is supported, including mixed Persian–English pages. Upload images or PDFs and download a searchable PDF, Word, Markdown, JSON, and Unicode text.

Author: [Alireza Aminzadeh](https://huggingface.co/alirezaaminzadeh)

## What you get

- Images (`JPG`, `PNG`, `WEBP`, `BMP`, `TIFF`, `HEIC`) plus PDF
- Camera capture, clipboard, and multi-file batch
- Persian, English, mixed, or auto-detect
- Modes: document text, precise overlay, tables, Markdown, form fields, and compare
- Optional deskew/contrast enhancement and high-accuracy tiled reading
- Searchable PDF, `.txt`, `.md`, `.json`, `.docx`, and a ZIP of every file
- Editable text with rebuild (no second model run)
- Session history, bilingual UI, and a `/health` API
- No persistent storage of uploaded files

## Limits

- Up to `8` pages or files per job
- Maximum `25 MB` per file
- Longest side is resized to `1920` pixels (`2560` in high-accuracy mode)
- Best results on flat, sharp, well-lit pages

## Model

The Space runs [PaddleOCR-VL-1.6](https://huggingface.co/PaddlePaddle/PaddleOCR-VL-1.6) through Transformers on ZeroGPU. The official PaddlePaddle page-parser is not used, so the app stays compatible with Hugging Face ZeroGPU.

## Privacy

Uploads are written only to a temporary working directory for the current request and discarded after a short TTL.

## Local run

Copy `.env.example` to `.env`, then:

```bash
docker compose up --build
```

The app is available at [http://localhost](http://localhost).

Optional GPU runtime (NVIDIA Docker required):

```bash
docker compose -f docker-compose.yml -f docker-compose.gpu.yml up --build
```

Optional production locks in `.env`:

- `OCR_API_KEY` — required on every extract call when set
- `APP_AUTH_USER` / `APP_AUTH_PASSWORD` — HTTP basic auth for the UI

Monitor containers with `docker stats`.

## API

Gradio exposes named endpoints once the app is running:

- `/ocr` — extract text and files
- `/rebuild` — rebuild exports from edited text
- `/health` — JSON status and process metrics

Example:

```python
from gradio_client import Client, handle_file

client = Client("http://localhost")
status, warnings, text, preview, pdf, txt, md, json_file, docx, archive, job, history, table = client.predict(
    handle_file("scan.png"),
    None,
    "auto",
    "document",
    True,
    False,
    "",
    "en",
    [],
    api_name="/ocr",
)
```

## Tests

```bash
python -m pip install -r requirements-test.txt
python scripts/make_assets.py
python -m unittest discover -s tests -v
```

## License

Apache License 2.0. See `LICENSE` and `NOTICE`.