In plain language
It can read a photo of a receipt or a scanned form, not just typed text.
Example
A workflow sends a photo of a handwritten delivery note to a multimodal model and gets back the items and quantities as data.
Related terms
Large Language Model (LLM) · Optical Character Recognition (OCR) · Structured Output · Generative AI
Common misconception
Reading an image is not the same as reading it perfectly; blurry, rotated, or handwritten inputs still cause errors.
Implementation guidance
Send clear, well-lit inputs, ask for structured output, validate extracted values, and route low-confidence results to a person.