A modular image processing SDK comprises an API to receive API calls from third party software running on a portable device including a camera. SDK logic receives and processes commands and parameters received from the API that are based on the API calls received from the third party software. An annotation system performs image processing operations on a feed from the camera based on image processing instructions and parameters received by the annotation system from the SDK logic. The image processing is based at least in part on augmented reality content generator data (or AR content generators), user input and sensor data.