3 ms·

How accurate is this compared to simple text extraction? Can there be a hybrid text-vision mode for charts and graphs?