What is PaddleOCR Local?

PaddleOCR Local is a self-hosted document parsing workbench that converts PDFs, images, and Office documents (Word, PowerPoint) into editable Markdown. It provides a unified WebUI and CLI for five OCR models: PaddleOCR-VL 1.6, PP-OCRv6, OvisOCR2, HPD-Parsing, and NaviDC-OCR. The software runs entirely locally with no cloud uploads, supports model switching with automatic GPU memory release, and offers batch processing, progress tracking, and multiple export formats (Markdown, JSON, DOCX, searchable PDF, HTML, CSV/XLSX). It deploys via Docker Compose on Linux/Windows with NVIDIA GPU, has a CPU-only mode using PP-OCRv6, and provides native Apple Silicon support via MLX for OvisOCR2. The project includes an API token for access control and a FastAPI interface with OpenAPI documentation.

Should you self-host PaddleOCR Local?

PaddleOCR Local scores 44/100 on selfhost.directory (how scoring works) and ranks #35 of 46 in Document Management. It has 138 GitHub stars.

There was no stable release in the past 12 months, so check the repository's recent activity before you rely on it.

Setup is advanced: plan for several services and some ongoing maintenance. It is Apache-2.0-licensed and written mostly in Python.

Its closest alternative on selfhost.directory is Reactive Resume (87/100, 44 977 stars). It scores higher overall, so compare the two before you pick. See PaddleOCR Local vs Reactive Resume side by side.

Features

Apple Silicon / MLX support
Batch processing and CLI
Bilingual UI (Chinese/English)
CPU-only deployment
External OCR engine integration
FastAPI with OpenAPI
Five OCR models unified
GPU memory pre-check
Model switching with GPU memory release
Multiple Export Formats
Office document support
PDF to Markdown conversion
Side-by-side original and result view
Table and formula rendering
Task history and search

How to configure PaddleOCR Local

API_IMAGE_TAG_SUFFIXlatest-nvidia-gpu-offlineDocker image tag suffix for API services
Docker image tag suffix for API services
values: Image tag suffix
HPD_PARSING_MODEL_NAMEPaddlePaddle/HPD-ParsingHPD-Parsing model name
HPD-Parsing model name
values: Model identifier
NAVIDC_OCR_MODEL_NAMEStarDoc-AI/NaviDC-OCRNaviDC-OCR model name
NaviDC-OCR model name
values: Model identifier
OVISOCR2_MODEL_NAMEATH-MaaS/OvisOCR2OvisOCR2 model name
OvisOCR2 model name
values: Model identifier
PADDLEOCR_VL_MODEL_NAMEPaddleOCR-VL-1.6-0.9BPaddleOCR-VL model name
PaddleOCR-VL model name
values: Model identifier
PANDOCR_ACTIVE_MODEL_ON_STARTpaddleocr-vl-1.6Model to activate on startup
Model to activate on startup
values: Model ID from catalog
PANDOCR_API_TOKENAPI token for WebUI access control
API token for WebUI access control
values: Any string
PANDOCR_CONTROLLER_OCR_LEASE_TTL_SECONDS0OCR lease TTL in seconds (0 = no expiry)
OCR lease TTL in seconds (0 = no expiry)
values: Integer seconds
PANDOCR_CORS_ORIGINShttp://localhost:8000,http://127.0.0.1:8000Allowed CORS origins for WebUI
Allowed CORS origins for WebUI
values: Comma-separated URLs
PANDOCR_CPU_PORT8000Port for CPU deployment WebUI
Port for CPU deployment WebUI
values: Integer port
PANDOCR_ENABLE_API_DOCS0Enable FastAPI /docs endpoint
Enable FastAPI /docs endpoint
values: 0 or 1
PANDOCR_ENFORCE_ORIGIN_CHECK1Enforce origin check for cross-site requests
Enforce origin check for cross-site requests
values: 0 or 1
PANDOCR_GPU_DEVICE_ID0GPU device ID to use
GPU device ID to use
values: Integer
PANDOCR_GPU_RELEASE_TIMEOUT60GPU memory release timeout in seconds
GPU memory release timeout in seconds
values: Integer seconds
PANDOCR_MAX_CONCURRENT_OCR1Maximum concurrent OCR requests
Maximum concurrent OCR requests
values: Integer
PANDOCR_MAX_UPLOAD_MB512Maximum upload size in MB
Maximum upload size in MB
values: Integer
PANDOCR_MODEL_CATALOGpaddleocr-vl-1.6,pp-ocrv6,ovisocr2,hpd-parsing,navidc-ocrComma-separated list of enabled models
Comma-separated list of enabled models
values: Model IDs
PANDOCR_MODEL_CONTROLdockerModel control mode: docker, none, or external
Model control mode: docker, none, or external
values: docker, none, external
PANDOCR_MODEL_SWITCH_TIMEOUT1200Model switch timeout in seconds
Model switch timeout in seconds
values: Integer seconds
PPOCR_V6_MODEL_NAMEPP-OCRv6_mediumPP-OCRv6 model name
PP-OCRv6 model name
values: Model identifier
VLM_BACKENDvllmVLM backend for PaddleOCR-VL
VLM backend for PaddleOCR-VL
values: vllm
VLM_IMAGE_TAG_SUFFIXlatest-nvidia-gpu-offlineDocker image tag suffix for VLM services
Docker image tag suffix for VLM services
values: Image tag suffix