Abstract
The paper introduces ORAL2008, a new fully balanced 1-million corpus of authentic spoken Czech covering the whole of Bohemia. It describes its design fundamentals in the context of other spoken Czech corpora and the automatic balancing procedure.