← Latest products

We built an OCR server that can process 270 dense images/s on a 5090

TurboOCR, >200 img/s OmnidocBench. TensorRT FP16, PP-OCRv6, HTTP + gRPC.

Open source

1,153 stars · GitHubexternal site

Visit github.comexternal site

Reactions (1)

Similar products

by description similarity

  1. BetterOCR combines and corrects multiple OCR engines with an LLM

    github.com

    🔍 Better text detection by combining multiple OCR engines (EasyOCR, Tesseract, and Pororo) with 🧠 LLM.

    Hacker News63 pts

  2. OpenCV over WebRTC (in Go)

    github.com

    Examples of WebRTC applications that are large, or use 3rd party libraries

    Hacker News154 pts

  3. Irpapers

    github.com

    Visual embeddings vs. OCR trade-offs in scientific PDFs

    Hacker News5 pts

  4. Open-source 3B param model better than Mistral OCR

    nanonets.com

    Our open image-to-markdown OCR model — turning documents into structured text with tables, equations, and reading order preserved.

    Hacker News9 pts

  5. I embedded 685M public texts in 32 minutes (on 8x A100, Rust, TensorRT)

    github.com

    Fast self-hosted embedding engine for search, RAG, and reindexing workloads on NVIDIA GPUs. Built in Rust + TensorRT for teams that care about scale, cost, and control.

    Hacker News7 pts