Text Titan II

Text Titan II is the successor to Text Titan I and our most accurate general-purpose model to date. It transcribes both printed and handwritten documents across the major Latin-script languages, including German, English, French and Dutch, and is designed to perform well on the kind of material processed on Transkribus every day. Compared with Text Titan I, it reduces character error rates considerably across languages and document types, with the strongest improvements on handwritten material. As a single multilingual model, it removes the need to choose a language-specific model in most cases, while continuing to match or exceed our dedicated language models on their own benchmarks. We recommend Text Titan II as the default starting point for most Latin-script documents. For specialised scripts, historical hands, or languages outside its coverage, a dedicated model may still give better results.