Ah yes, these look a little “out of distribution” (not reflected in the training data) for the moment. But we’re adding more examples of medieval Latin / book hand for the next version of the model, so hopefully Leo will have a better shot at tackling these soon.
Leo did the same thing for one of my images. It was able to transcribe the text “Cornelia died Oct 15-19” but then for some reason it just kept repeating that same line over and over again instead of transcribing the rest of the document.
