Skip to main content
Retto Logo

Welcome to Retto

Retto is a high-performance OCR (Optical Character Recognition) library powered by PaddleOCR models, built in Rust for maximum performance on desktop and WebAssembly platforms.

Key Features

Multi-Platform Support

Run on CPU, CUDA, DirectML, or WebAssembly with the same API

High Performance

Rust-powered inference with optimized image processing pipelines

Easy Model Loading

Load models from local files, HuggingFace Hub, or embed them directly

Streaming Inference

Get real-time callbacks for detection, classification, and recognition stages

Get Started

Installation

Install Retto for Rust, CLI, or WebAssembly

Quickstart

Run your first OCR in under 5 minutes

API Reference

Explore the complete API documentation

Use Cases

Retto is perfect for:
  • Desktop Applications: Build native OCR tools with Rust for maximum performance
  • CLI Tools: Batch process images from the command line
  • Web Applications: Run OCR directly in the browser with WebAssembly
  • Document Processing: Extract text from images, receipts, invoices, and more
  • Accessibility Tools: Make visual content accessible through text extraction

What Makes Retto Different?

Retto combines the power of PaddleOCR models with Rust’s performance and safety guarantees, delivering OCR capabilities across desktop and web platforms with a unified API.

Rust Library

Use Retto as a Rust crate with full type safety and performance

CLI Tool

Process images from the command line with batch support

WebAssembly

Run OCR in the browser with our npm package

Flexible Backends

Choose the right backend for your platform and requirements

Build docs developers (and LLMs) love