Thanks! I used a jpeg of the source document page and a written document transcript excerpt, fed into NightCafe's ai engine. I haven't explored VQGAN+Clip much, but I should: )