Voice user interface (VUI)
A voice user interface (VUI) is an interface that lets a person interact with a system by speaking and listening rather than relying only on visual screens, typing, or physical controls.
A VUI defines the user experience around the underlying speech technology. Speech recognition captures what the person says, language processing interprets it, and speech synthesis or recorded audio delivers a response. The interface layer determines which requests are supported, how the system asks questions, what it says when input is unclear, and how users move forward, correct a mistake, or leave the interaction.
Voice interfaces range from command-based systems to open-ended conversational agents. A command interface may accept a small vocabulary such as “start,” “stop,” or a number. A conversational VUI can handle varied phrasing and maintain context over several turns. Both can be appropriate. The right design depends on the task, the cost of misunderstanding, and how much freedom helps the user.
Phone calls are entirely voice-first, so the interface must make choices visible through sound. A screen can display several options at once; a caller has to remember options in sequence. Prompts should therefore be short, concrete, and ordered by importance. The system should avoid long lists, confirm details that are difficult to hear, and give callers a clear way to repeat, correct, transfer, or go back.
Timing is also part of VUI design. If the system waits too long, the caller may think it failed. If it responds too quickly, it may cut off a pause that was part of the caller’s sentence. Endpointing, turn-taking, and barge-in behavior shape the conversation as much as the words do. Designers should test these behaviors with natural speech rather than asking participants to follow a fixed script.
A good VUI is measured by whether people can understand the system and complete the task with reasonable effort. Useful tests include interruptions, background noise, ambiguous answers, corrections, silence, and requests outside scope. The interface should explain a problem in plain language, offer a practical recovery path, and make human help available when the automated path is no longer appropriate.