How does a voice user interface work?

Every VUI runs the same loop: capture speech, recognise it, work out what the person wants, act or answer, and speak the response. Designing one means deciding what the system says, how it confirms what it heard, how it recovers when it doesn’t understand and how people know when to talk.

Standards shaped the field. The W3C’s Voice Browser Working Group produced VoiceXML 2.0, which brought web development to interactive voice response (IVR) systems with prompts, grammars, timeouts and barge-in, and SSML to control synthetic speech. On the web, the Web Speech API lets a page add speech recognition and speech synthesis.

Older VUIs understood fixed phrases, such as “say billing or technical support”. Many newer ones use large language models, so people can say what they need in their own words.

Example: a VUI on a website can let a visitor speak a question and read captions of the answer at the same time, combining voice with the screen.

Sources

What makes a voice user interface easy to use?

Listeners can’t scan back the way readers can, so voice needs its own design rules:

  • Say what it is and what it can help with, briefly, at the start.
  • Keep each answer short enough to remember, and give steps one at a time.
  • Make it clear when it’s listening, and let people interrupt.
  • Confirm important details, such as names and numbers, and recover politely from mishearing.
  • Offer another way in, such as captions or typing, for people who can’t or would rather not speak.

How does Kawl Now’s voice interface work?

Kawl Now’s interface is part voice and part screen. Before the call, the visitor types what they need, in up to 1,000 characters, and reads a notice that they’re speaking with AI, that a transcript will be saved and that calls last up to 10 minutes.

During the call the conversation is spoken, with on-screen buttons to mute, show live captions and end the call, and a timer against the limit. If the browser blocks audio, an Enable audio button appears.

There’s no wake word or hands-free start: a call begins only when the visitor clicks the widget and allows the microphone. Agents greet callers in English, and there’s no typed chat, so the captions are the text view of the call.

Common questions

What is the difference between a VUI and a GUI?

A graphical user interface (GUI) is used by looking and pointing: menus, buttons and forms on a screen. A voice user interface is used by speaking and listening. Many products combine the two, such as a voice agent that shows captions and buttons while you talk.

What does VUI stand for?

VUI stands for voice user interface: the spoken side of a product, covering what it says, how it listens and how people move through a conversation with it.