Why voice agents borrow the video-call layout
Talking to a machine is still strange. Put it inside the layout everyone knows from FaceTime and Zoom, agent as the main character, you as the small self-view, and the strangeness melts into muscle memory.
Try the live demoVoice is the most natural interface humans have, and somehow talking to a voice assistant has always felt the least natural thing in tech. Part of the problem is that there is nowhere to look. You speak into a void and a disembodied voice answers. There is no presence, no focal point, no sense of who or what you are talking to.
The fix turned out to be a layout we already had. The video call. For fifteen years we have been training ourselves to talk to a screen with a face in the main frame and a little thumbnail of ourselves in the corner. Drop an AI agent into that exact shell and the interaction inherits all of that comfort for free.
The clever part is that none of it is new. The agent is the main character, the user is the self-view, the controls sit where they always sit. The familiarity is the feature. A brand-new way of interacting arrives wearing clothes the user has worn a thousand times.
Bare voice, or the call frame
Switch between a disembodied voice and the video-call frame, swap the orb for an avatar, and tap to make the agent speak. Notice how much more present the framed version feels.
Voice with no frame feels like talking to a wall. There is nowhere to look, no sense of presence, and no familiar shell to absorb the novelty of speaking to a machine. An abstract orb sidesteps the uncanny valley and reads as honestly artificial, while still giving the eye a focal point.
What the pattern looks like done well
Six rules for wrapping a voice or video agent in a frame people trust.
Borrow a layout the user already knows
The video-call frame is not decoration. It imports years of muscle memory from FaceTime and Zoom, so a brand-new interaction (talking to an AI) arrives inside a familiar shell.
Make the agent the main character
The agent takes the main frame; the user sits in a small self-view. This mirrors a call with a person and quietly frames the agent as a presence you are meeting, not a tool you are operating.
Keep the self-view, even though it is just you
The little corner tile of yourself is doing real work: it confirms the system can see and hear you, and it completes the call metaphor. Removing it makes the interaction feel one-sided and uncertain.
Choose orb or avatar deliberately
An abstract orb is honest about being artificial and dodges the uncanny valley. A realistic avatar adds presence but raises the bar: it must be responsive and lifelike, or it unsettles. Pick for the context, not the novelty.
Show presence and state, not just audio
Speaking animations, a connection indicator, and a listening state give the agent visible aliveness. Silence with no visual signal reads as broken, even when the audio is fine.
Keep the controls where calls keep them
Mute and end-call belong at the bottom centre, where every video app has trained users to find them. Familiar control placement lowers the cognitive cost of a strange new medium.
Frequently asked questions
Don’t just read this. Put it to work.
The whole series is distilled into one Markdown file: every pattern, the do and don’t rules, and how well each is evidenced. Download it into your project, or paste the link into any chat with your agent and tell it to improve your agent UX. It’s free, no sign-up, no attribution required.
Use these Agent UX principles to review and improve our agent's interface: https://p0stman.com/agent-ux/agent-ux-principles.md
We have already shipped this one
The Zee video agent on p0stman.com runs the exact pattern in this article: a real-time voice agent in a familiar call frame, agent as the main character, user self-view in the corner. If you want a voice or video agent that people are actually comfortable talking to, that is the job we do.