One of the acknowledged impediments to the widespread use of speech interfaces is the portability problem, namely the considerable amount of labor, data and expertise needed to develop such interfaces in new domains. Under the Universal Speech Interface (USI) project, we have designed unified look-and-feel speech interfaces that employ semi-structured interaction and thus obviate the need for data collection. More importantly, the unified structure of USI-compliant interfaces makes possible the automatic generation of new interfaces from a terse, high-level specification. In this paper, we describe an application generator and accompanying toolkit that allow even non-programmers to generate and use fully functional speech interfaces to their chosen database in less than 15 minutes.