TurboOCR

GPU document parser in C++/TensorRT: PP-OCRv6 OCR, layout, tables and formulas to Markdown at 200-559 img/s on one GPU — no VLM, HTTP + gRPC, one-line Docker deploy.

894 88 C++ MITupdated 3 days ago

ocr

View on GitHubDispute this mapping →

Curator's take

The throughput bet: a classic multi-stage pipeline on TensorRT instead of a VLM — ~20 pages/s structured parsing where VLM parsers run ~1. Sharp on forms and receipts. When NOT: Linux + NVIDIA Turing+ only (Metal/ROCm still cooking), first start compiles engines (up to an hour on older cards), and benchmarks are self-published on its own harness — for messy scans and handwriting a VLM parser may still out-read it.

Mapped by ShipWithAI editors · links verified

README.md

_{🧪 Apple Metal and Intel OpenVINO backends are in testing, with AMD ROCm in development. NVIDIA is the only backend shipped today.}

English | 简体中文

The fastest GPU document parser — OCR · layout · tables · formulas → Markdown, at 200–559 images/s on one GPU.
C++ / CUDA / TensorRT / PP-OCRv6 — Linux + NVIDIA GPU

🎉 v3.0 — now powered by PP-OCRv6

_{New medium / small / tiny tiers · higher accuracy · faster defaults · breaking changes}

⭐ Star TurboOCR on GitHub — it helps others (and agents) find it.

Quick Start · Accuracy · Benchmarks · Models · v3 changes · API · Docs

An extremely fast GPU document parser — not just OCR. PP-OCRv6 detection + recognition, plus layout, tables (→ HTML), formulas (→ LaTeX) and reading-order Markdown, the whole pipeline on a single multi-stream CUDA/TensorRT engine, locally (no VLM), behind HTTP and gRPC. Whole-page OCR runs at up to 559 images/s on receipts (one RTX 5090), and full structured parsing (layout + tables + formulas) at ~20 pages/s — where VLM document parsers like PaddleOCR-VL run ~1 page/s. On forms and receipts it is accurate and 15–90× faster than classic OCR engines.

🚀 **559 img/s (receipts) · 520 (forms) · 200+ (dense docs) on one RTX 5090 — fastest by default
🎯 Accurate on forms & receipts — competitive with PaddleOCR-VL, PaddleOCR-Python, RapidOCR, EasyOCR and Tesseract (benchmarks)
🧠 PP-OCRv6 — one model covers Latin + Chinese + Japanese; pick tiny (default) / small / medium
🌐 More scripts — Arabic, Cyrillic, Korean, Thai, Greek via retained PP-OCRv5 recognizers
📄 **PDF n

Continue your stack

What teams reach for next — and why each earns a place beside TurboOCR. Ranked by curator confidence.

Pairs well with

pdf-inspector

Firecrawl's Rust PDF triage: classifies text-based vs scanned in ~10-50ms, extracts positioned text and clean Markdown without OCR — routing the ~54% of PDFs that never needed a model.

Why it fitspdf-inspector triages text vs scanned in ~10-50ms; route only the scanned share to TurboOCR's GPU and skip OCR for the rest.

Alternative to See all + compare →

MinerU

Heavyweight document-to-markdown/JSON parser — PDFs plus Office (docx/pptx/xlsx) through layout analysis and OCR into LLM-ready output for RAG and agentic pipelines. 73k stars, self-hostable.

Why it fitsSame job — documents to LLM-ready Markdown. MinerU is the heavyweight Python pipeline with Office support; TurboOCR trades breadth for 200+ img/s on one GPU.

olmocr

Open toolkit that linearizes messy PDFs — scans, tables, equations, handwriting — into clean ordered Markdown with a self-hosted vision-language model. Built for LLM training data and RAG ingestion.

Why it fitsolmOCR linearizes messy PDFs with a self-hosted VLM; TurboOCR refuses the VLM entirely and wins ~20x on throughput. Pick by corpus: pristine forms → TurboOCR, chaotic scans → VLM.