At present, many voice input systems are unable to consistently achieve accurate recognition in practical environments. This paper describes issues and some solutions concerning voice input applications from two different standpoints of a speech recognition researcher and an application designer. The authors point out that both the robustness in speech recognition and the well-designed user interface are particularly important for voice input applications to successfully incorporate speech recognition and to complete user's demands. Some practical applications, such as voice-activated telephones, ticket vending machines, and elevators, are also discussed, and two important items of the multimodality and well-designed prompts are emphasized.