Seems to do an OK job:

https://g.co/gemini/share/e173d18d1d80

This is a random image from Twitter with no transcript or English translation provided, so it's not going to be in the training data.

Right, it can do modern writing but anything older than a century ( church records and census)and it produces garbage. Yandex Archives figured that out and have CER in a single digit but they have the resources to collect immense data for training. I'm slowly building a dataset for finetuning TROCR model and the best it can do is CER 18% ... which is sort of readable.