OCR
Turning an image of text into text you can select and search.
Optical character recognition reads text out of an image: a scanned invoice, a photo of a whiteboard, a screenshot of a form. It produces plain text that can be searched, copied, and passed to a model as context.
Modern OCR is very good on clean, printed documents and much weaker on the rest: handwriting, low contrast, skewed photos, dense tables where the column order is ambiguous. It also silently normalises what it reads, so a misread digit becomes a wrong number with no warning.
For a document workflow, the practical choice is to avoid the image entirely when the file exists. A PDF with a text layer gives you exact characters, correct reading order, and no OCR step to debug. Reach for OCR when the source genuinely is a picture, and treat its output as a draft that a human or a checksum should confirm.
Related terms
Vision Model
A model that reads images, turning them into something a language model can reason over.
Multimodal
A model that handles more than one kind of input or output: text, image, audio, video.
Vision Model
A model that reads images, turning them into something a language model can reason over.
Structured Output
Making the model answer in a fixed shape a program can read.
RAG
Fetch the relevant documents first, then let the model answer using them.
Want this working inside your team?
It starts with one conversation about your workflow.