This article focuses on the learning that is enabled while a primary school child makes a story using multimodal software. This child is diagnosed with autism. The aim is to use a cultural-historical framework to carry out an in-depth analysis of a process of learning with action as a unit of analysis. The article is based on a collaborative research project in which assignments on narration concurrently served as experiments with cultural tools. The child's learning process in one assignment is analyzed sequentially, highlighting changes in the action, based on data from videotape. The child's decision to record audio was a turning point that expanded the formal learning and animated the idea of making the story in a desired multimodal format. The multifaceted quality and the dialectics revealed in the process are vital to achieve expansive action and learning. The combination of a generative narrative model and the multimodal affordances is crucial for actualizing the potential in the research approach to accomplish an analysis of the learner's changes "in" action. Thus, the study indicates the relevance of using a cultural-historical framework in research and practice that focuses on processes of learning and processes of becoming among diverse learners.