Transfer functions play a critical role in feature detection through direct volume rendering in volumetric scalar fields. Because of the inherent difficulties of exploratory visualization, assisting the user in transfer-function specification is still an important research area. In particular, the disparate nature of simulated and measured unstructured volumetric data necessitates more flexibility in the specification process than existing tools provide. Here, the authors introduce a framework that combines elements of existing techniques and introduces new features to handle the diversity of volumetric data sets. The authors discuss a range-mapping technique for assisting in transfer-function specification for high-dynamic range data, an algorithm for combining multiple transfer functions into ensembles that allow multiple specification techniques to be used collectively, and an intuitive method for specifying transfer functions for multiple classes of time-varying data.