ocr

This paper proposes an agentic framework combining rules and LLMs for embedding and annotating document layouts in plant science, demonstrating scalable trait extraction with improved annotation coverage.

Caplio is a Mac application that automatically organizes screenshots and images into a searchable private library using on-device OCR.

TeleOCR is a lightweight open-source Vision-Language Model for document parsing, achieving state-of-the-art performance on both digital and camera-captured documents using techniques like Multi-node Consensus Voting.

TeleOCR is a lightweight open-source Vision-Language Model designed for document parsing, unifying digital and camera-captured documents with state-of-the-art performance on benchmarks.

Mistral AI announces OCR 4.1, an updated OCR service with native paragraph-level bounding box extraction, structural block labels, and block-level confidence scores, now in public preview.

NaviDC-OCR is a unified vision-language framework that improves document parsing accuracy through deformation-aware learning and adaptive sampling, achieving state-of-the-art results on multiple benchmarks.

LiquidAI releases LFM2.5-VL-3B, a 3B multimodal model for on-device deployment with improved OCR, grounding, and efficient inference, available in multiple formats including GGUF, ONNX, and MLX.

TongGuOCR is a layout-aware and token-augmented OCR framework for Chinese historical documents, achieving state-of-the-art results on two benchmarks, including 93.76 AR on the challenging M5HisDoc benchmark.

The author shares frustration with duct-taping data prep tools for AI agents and proposes a simpler approach: upload raw files, describe the desired output in plain English, and receive cleaned, structured data ready for agents.

A user shares hands-on impressions of a thin, light Android-based device featuring microphone, speaker, OCR, text-to-speech, and a ChatGPT 5.2 chat, noting the only clear drawback is its lack of color.

NVIDIA releases Nemotron Parse 2.0, a document image parsing model that converts scanned PDFs and images into structured text with layout, bounding boxes, and reading order, adding multilingual OCR improvements and chart-aware parsing.

A benchmark comparing 8 PDF parsers across 14 capabilities, finding Chandra the most accurate while noting trade-offs like speed; LightOnOCR-1B impresses for its size but hallucinates on illegible text.

Xberg is a local content intelligence framework (Rust core, MIT) that extracts text from 101 document formats via MCP server or CLI, reconstructs reading order and tables, and chunks for context windows—all on-device for AI agents.

A developer benchmarks three PDF-parsing models (MinerU, Granite-Docling, PaddleOCR-VL) across six document types and twelve capabilities, finding MinerU drops footers unless its markdown is rebuilt while Granite-Docling outputs cleaner native markdown tables.

A GitHub project, llm_aided_ocr, performs OCR on scanned PDFs and uses LLMs to correct and improve the extracted text.

Klartext is a free, privacy-focused web tool that converts PDFs, scans, photos, Word, Excel, and PowerPoint files into Markdown (plus JSON) entirely on its own server, with automatic deletion after 24 hours and no external AI/OCR services.

Space OCR is an OCR tool that checks its own answers, available as an app or an API.

This paper presents IDP AutoOpt, an autonomous LLM agent that optimizes intelligent document processing pipeline configurations, matching or exceeding human-expert accuracy at lower cost and reducing configuration time from weeks to under two hours.

A fully offline AI reads over 4,000 pages of declassified UFO files using OCR and vector database, providing cited answers locally without cloud or API keys.

A developer describes building a PDF triage tool that inspects the PDF structure to determine if it contains a text layer or is made of scanned images, preventing silent empty extraction failures, and shares pitfalls around false text detection and phantom tables.