OCR / document text extraction
RapidOCR is a lightweight, open-source OCR toolkit that converts PaddleOCR models to run on ONNX Runtime/OpenVINO/MNN/TensorRT and provides inference APIs in multiple programming languages.
Solves the same problem with a different architecture or at a different layer. Expect to rewrite the integration.
Choose this original PaddleOCR framework when you need model training/fine-tuning and native PaddlePaddle support, at the cost of heavier dependencies.
Turn any PDF or image document into structured data for your AI. A powerful, lightweight OCR toolkit that bridges the gap between images/PDFs and LLMs. Supports 100+ languages.
Choose when you can use vision language models for layout-aware, document-level OCR instead of a local neural OCR pipeline.
OCR & Document Extraction using vision models
Choose for a research-grade document image parsing model with heterogeneous anchor prompting if you accept working from an unreleased codebase.
The official repo for “Dolphin: Document Image Parsing via Heterogeneous Anchor Prompting”, ACL, 2025.
Choose when you need an end-to-end vision-language OCR model that reads images without separately detecting text boxes.
GLM-OCR: Accurate × Fast × Comprehensive
Choose for a maintained, pluggable OCR engine with PyTorch/TensorFlow backends and an active community.
docTR (Document Text Recognition) - a seamless, high-performing & accessible library for OCR-related tasks powered by Deep Learning. Ongoing development and maintenance by t2k.
Choose if you want a broad toolbox of classic OCR algorithms for experimentation or research in OpenMMLab.
OpenMMLab Text Detection, Recognition and Understanding Toolbox
Choose for a lightweight tool that converts images/PDFs to Markdown, including text, layout, tables, and math formulas.
An Open-Source Python3 tool with SMALL models for recognizing layouts, tables, math formulas (LaTeX), and text in images, converting them into Markdown format. A free alternative to Mathpix, empowering seamless conversion of visual content into text-based representations. 80+ languages are supported.
Choose for advanced PDF-to-Markdown with complex layouts, tables, and cross-page content if you do not need a tiny embedded library.
OCRFlux is a lightweight yet powerful multimodal toolkit that significantly advances PDF-to-Markdown conversion, excelling in complex layout handling, complicated table parsing and cross-page content merging.
Choose for a PyTorch-based OCR library with broad language/writing-script coverage and a simple API.
Ready-to-use OCR with 80+ supported languages and all popular writing scripts including Latin, Chinese, Arabic, Devanagari, Cyrillic and etc.
Choose if you need OCR along with layout analysis, reading order, and table recognition in one unified pipeline.
OCR, layout analysis, reading order, table recognition in 90+ languages
An earlier or more limited way to do the job. Still the right call when you need something small, proven, or CPU-only.
Choose only for legacy Apple Swift projects needing an offline OCR, despite outdated model quality and no maintenance.
Fast and simple OCR library written in Swift
Choose for a simple Keras CRAFT+CRNN pipeline if your workload is Latin-text OCR and you want minimal dependencies.
A packaged and flexible version of the CRAFT text detector and Keras CRNN recognition model.
Does the same job, but ships as an app. Useful to a person, not swappable into a codebase.
A TensorRT-accelerated OCR server with HTTP/gRPC endpoints; choose if you need high-throughput deployment as an application rather than an embeddable library.
TurboOCR, >200 img/s OmnidocBench. TensorRT FP16, PP-OCRv6, HTTP + gRPC
These projects were analysed and named RapidOCR among their alternatives. The relationship is not symmetric — how RapidOCR rates them is a separate judgement, made when RapidOCR is analysed in its own right.
calls RapidOCR “Choose when you need ONNX-based OCR that runs on CPU/edge devices with no PyTorch dependency.”
Ready-to-use OCR with 80+ supported languages and all popular writing scripts including Latin, Chinese, Arabic, Devanagari, Cyrillic and etc.
calls RapidOCR “Apache-2.0, actively maintained, supports ONNX/OpenVINO/TensorRT backends; pick for broader platform integration and permissive licensing.”
超轻量级中文ocr，支持竖排文字识别, 支持ncnn、mnn、tnn推理 ( dbnet(1.8M) + crnn(2.5M) + anglenet(378KB)) 总模型仅4.7M
calls RapidOCR “ONNX-based OCR toolkit with fast CPU inference and multiple language bindings.”
OCR, layout analysis, reading order, table recognition in 90+ languages