vision API
An API endpoint that takes an image and returns an analysis—caption, object detection, OCR, reasoning about visual content.
Think of it like
Giving a description of a painting to a blind person; the model translates pixels to words.
Example
Send a photo of a receipt to Claude's vision API. Get back line items, total, vendor, date—all extracted and structured.
How it actually works
Vision APIs work because large models are trained on both images and text. They can answer 'what's in this image,' 'describe the text,' 'count the objects,' etc. Multimodal models like GPT-4V and Claude 3 handle this natively. Vision adds a dimension to automation (no more manual data entry from PDFs).
For product teams
Enables automation of visual tasks (invoice processing, content moderation); unblocks new user workflows.
For engineers
Takes images as base64 or URLs; streaming not yet common; usually slower than text.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome