Comprehending results from 3D CFD simulation is a difficult task. In this paper, we present a semantics-based approach to featurebased volume rendering of 3D flow data. We make use of interactive feature specification to acquire derived data about what is most interesting to the user. This process results in so-called degree-ofinterest (DOI) values, associated with the original data items. This information is then used during the visualization mapping to allow for visualization of flow semantics. We present three different approaches in this paper: (a) isosurfacing the degree of interest – the result is a triplet (or quintuplet) of iso-surfaces representing levels of interest; (b) feature volumes – volume rendering is used to depict 3D distributions of DOI values; and (c) interest-based seeding of streamlines, resulting in reduced clutter while focusing on the most interesting parts of the 3D flow. We utilize fast volume rendering (RTVR) for real-time viewing, also providing ...