Vision is Apple's computer vision framework, a library in Apple's operating systems that analyzes still images and video frames with pretrained machine learning models. Computer vision is the field of software that extracts information from images.
Every analysis follows the same three steps: the app creates a request for one kind of analysis, performs it on an image or a video frame, and reads the resulting observations, which hold details such as the recognized text, a confidence score and the bounding box (the rectangle around the found item). Vision recognizes text in 26 languages, finds tables and document structure, detects barcodes and QR codes, finds faces and facial features, isolates people and foreground objects from the background, tracks body poses of people and animals, classifies images and measures image quality. A CoreMLRequest runs a developer's own Core ML model, and OCR and barcode tools can be given to Apple's Foundation Models language model. Since iOS 18 there is a Swift-only API next to the original one.
The framework has no separate price; distributing apps requires the paid Apple Developer Program.