ReTV (http://retv-project.eu/) is a Horizon 2020 funded research project that gathers together computer scientists, broadcasters, interactive TV companies and audiovisual archives from across Europe with the goal to unlock the potential of audiovisual media assets in the digital realm and bring the legacy of broadcaster archives to the 21st-century audiences. ReTV is building an innovative infrastructure that uses machine learning techniques to enhance and tailor audiovisual content for online publication and repurposing. The project is currently developing tools that segment video materials from media management systems into multiple granularities and automatically annotate them with concepts depicted in the footage. During our presentation, we will demonstrate these AI-based video annotation algorithms, more specifically a web-based concept-based video annotation service running on real video examples. The service segments any submitted video to its constituent parts, i.e. shots and scenes, and evaluates a few hundred visual concept detectors for each video shot, generating shot-level concept-based annotations.