The Text Titan I bis

Text Titan I bis is a general-purpose text recognition model in Transkribus, designed to deliver strong results on a wide variety of Latin scripts without the need for custom training. It works well with both printed and handwritten texts and performs reliably even on less common writing styles. The model is built on a carefully selected and balanced training set, representing a wide range of historical and modern documents. This foundation allows it to generalise well and handle unexpected inputs with confidence. In evaluation, Text Titan I bis shows an average improvement of nearly 2 percentage points in character error rate compared to its predecessor, with the largest gains seen in English, unusual scripts, and handwritten material. Users working with diverse Latin script collections can expect robust recognition quality out of the box, making Text Titan I bis a suitable choice for both individual projects and large-scale document processing.