We only have limited information on how cinematic rendering is implemented in various commercial software (only what we can deduce from demos, presentations, etc.) but probably they follow a similar approach (it is fairly trivial).
Note that if segmentation result is a set of non-overlapping segments then it is not always necessary to use multivolume rendering, but RGB color channels can be generated from the segmentation result, which can be combined with alpha channel can be computed using a scalar opacity transfer function:
(see Color Volume Rendering: vtkDataArray Indexing - #5 by pieper)