GLM-OCR is a free online OCR tool that uses a 0.9B‑parameter AI model to extract text from images and PDFs, supporting tables, formulas, handwriting, and 8+ languages with state‑of‑the‑art accuracy.
What is GLM-OCR?
GLM-OCR is an open‑source (Apache‑2.0) online OCR tool that converts images and PDFs into editable text, Markdown, LaTeX, or structured JSON. It uses a CogViT encoder paired with a GLM decoder to handle complex layouts, scanned documents, photos, and screenshots. The tool is developed by the zai‑org team and is available as a free web app, a cloud API, and local deployment options.
Key Features
- Image to Text Converter – Extract text from photos, screenshots, and scanned documents with 99.9% accuracy on clean text; supports handwriting, seals, code, and special characters.
- Table Recognition & Extraction – Recognizes complex table structures and outputs them as Markdown or structured data suitable for spreadsheet import.
- Formula Recognition to LaTeX – Converts printed and handwritten mathematical formulas into LaTeX notation, achieving a 96.5 score on the UniMERNet benchmark.
- Multilingual OCR – Processes documents in 8+ languages (English, Chinese, Japanese, Korean, French, German, Spanish, Russian) with native‑level accuracy.
- Developer OCR API – REST API with JSON output for integration into custom workflows; priced at $0.99 per million tokens for cloud access.
- Flexible Deployment – Available via cloud API, Ollama, vLLM, SGLang, Hugging Face Transformers, or Docker for local or production use.
- PDF Text Extraction – Extracts text from PDF files up to 10 MB, preserving layout and parsing financial reports, academic papers, and business documents.
Who is it for?
- Academic researchers – Digitize archives, handwritten notes, and scientific papers, extracting text with preserved citations and LaTeX formulas.
- Financial analysts – Convert scanned financial statements and invoices into structured data ready for spreadsheet or database import.
- Legal professionals – Process contracts and case files, identifying clauses and structural hierarchy to streamline document review.
- Developers – Build OCR‑powered apps using the free API (or deploy locally) with straightforward JSON output and SDK support for Python, Node.js, and REST.
What can you do with GLM-OCR?
- Digitize research archives – Upload old papers or handwritten notebooks and receive searchable text with mathematical formulas converted to LaTeX.
- Parse financial documents – Extract tables from scanned quarterly reports and get structured data for analysis without manual re‑entry.
- Automate invoice processing – Use the API to extract key fields (dates, totals, line items) from receipts and invoices, outputting JSON for ERP integration.
- Extract code from screenshots – Grab code blocks from technical documentation images and paste them directly into your editor.
How does it work?
- Upload – Drag and drop an image or PDF (JPG, PNG, PDF; max 10 MB) into the free online interface.
- AI Processing – The CogViT encoder captures pixel details while the GLM decoder aligns visual features with language understanding, recognizing text, tables, and formulas.
- Get Results – Download the extracted text as plain text, Markdown, LaTeX, or JSON. Copy directly to clipboard or use the result in downstream tools.
Pricing
GLM-OCR is freemium. The online OCR tool at glm-ocr.com/ocr-tool is completely free to use. For developers, the cloud API costs $0.99 per million tokens. The model can also be self‑hosted at no cost via Ollama, vLLM, SGLang, Transformers, or Docker (model weights are open‑source under Apache‑2.0). Source code is on GitHub and pre‑trained weights on Hugging Face.
FAQ
