Comment by MattRogish

I do OCR of images, and that's exactly what I do. I take one big image and slice it into many smaller ones, and send those to the LLM. Perfect every time, unlike using the whole image which resulted in hot garbage.