How is it possible to convert images into text before ChatGPT fine-tuning?

I want to fine-tune the ChatGPT model for generating user guides according to UI mockups. The scenario is as follows: