Odiff: The "fastest image comparison library in the world" was rewritten in Zig (github.com)

ODiff, a high-performance native image comparison library originally written in OCaml, has been fully rewritten in Zig, delivering some of the fastest visual difference detections on the market. Optimized with SIMD instructions for SSE2, AVX2, AVX512, and NEON, ODiff can compare images across formats like PNG, JPEG, and TIFF within milliseconds, making it ideal for use cases involving screenshots, photos, and AI-generated images. Its support for cross-format comparison, anti-aliasing detection, region ignoring, and the use of the YIQ NTSC transmission algorithm for assessing visual differences sets it apart as a precise and versatile tool for visual regression testing. This rewrite in Zig not only boosts speed — benchmarks show ODiff outperforming popular tools like ImageMagick and Pixelmatch by up to 5-7x — but also maintains a controlled memory footprint and ensures 100% test coverage with backward compatibility. The library offers a simple CLI and lightweight Node.js bindings, enabling seamless integration into frontend testing pipelines and CI workflows. Projects like Cypress-ODiff, LostPixel, and Visual Regression Tracker have already adopted ODiff as a core engine, highlighting its practical utility for scalable, automated visual testing. For AI/ML practitioners and developers dealing with large-scale image evaluation tasks, ODiff's combination of multi-architecture SIMD optimization, cross-format support, and extensible API makes it a compelling choice to accelerate and enhance visual difference detection in machine learning model validation, UI testing, and beyond.