Pictures Found by What They Show

Pictures Described by What They Show

A photo, a drawing or a diagram inside a PDF often has no text in it, so reading it finds nothing. Opensolr gives such a picture a short description of what it shows, written by a model that runs on Opensolr's own servers, and stores it with the page. The page is then found by what the picture shows: for example, a search for "wind turbine" can find the page with the turbine photo.

When a picture gets a description

1. Opensolr first reads the picture for printed text: Text in scans and pictures.
2. Only a picture where no text was found is described.
3. The description becomes that picture's text on its page, in the index and for the chat bot.

It happens on every crawl, for the pictures in the PDFs found on your site, and on Data Ingestion, for the PDFs you send.

What it uses from your plan

• Descriptions use the AI of the plan of the account that owns the index. Each picture described uses part of one AI request; a picture described before is free.
• Without AI in the plan, or once this month's AI allowance is used up, the text of your documents is still read and indexed; only the descriptions are skipped.

More: What site search uses from your plan.

Picture meaning, elsewhere

• Search by image: a visitor's photo is turned into words for what it shows, and those words run the search.
• Your own code: describe pictures by API: Pictures and documents by API.