Multimodal models read images as well as text. A screenshot of an error, a chart or a form can go straight into a request.
What you need
- The Multimodal input guide.
The shape of it
Attach the image
Include the screenshot alongside your question in the request.
Ask a specific question
"What is the error message, and which line caused it?" gets a better answer than "what is this?".
The complete recipe is on upfyn.com: Read a screenshot.