October 11, 2026 · 5 min · personal agents · product · design
.mdYour personal agent deserves a face
Instinct, Muse, and Tab are making personal agents part of everyday life. Realtime Avatar helps developers give their own agent a face people can talk to.
A personal agent becomes more useful when you can think out loud with it. You can explain why a plan feels wrong, interrupt an assumption, or work through a decision before handing over the task. At Realtime Avatar, we see a place for a face in those conversations: a familiar character you can call when a message thread feels too narrow for what you need to say.
Imagine opening your agent before a busy day. It has your priorities and a few decisions waiting. You start talking through the first one, change your mind halfway through, and ask it to take a different approach. The useful part is the work it can do. The experience is the conversation you have while deciding what that work should be.
Instinct, Muse, and the personal-agent moment
Instinct makes the interaction familiar: you text or call an assistant that can operate a phone and computer. Its examples include ordering groceries, arranging appointments, and planning a weekend away. The invitation is to delegate something from your actual life through a conversation.
Meta introduced Muse on September 8, 2026, with a similar ambition at a much larger platform scale. Muse works through its app and WhatsApp, uses a dedicated virtual computer, and can continue working after you close the app. Meta describes an agent that remembers relevant details and returns when it needs your input.
Tab is another recent entrant built around messaging. Its site describes an assistant with its own computer and browser, able to work with documents, make phone calls, and bring results back into the same thread. Together, these products make an appealing promise: you can ask for help in ordinary language and return later to work that has moved forward.
For developers building personal agents, that raises an interesting design opportunity. An agent may handle dozens of tasks quietly in the background. The moments when someone wants to discuss a choice, explain a concern, or change direction deserve just as much attention as the automation itself.
A face for the moments that need a conversation
Meta has already made an explicit move in this direction. On September 23, it introduced Muse Realtime Avatar, describing technology that turns reference images into expressive characters for live conversations. Its announcement connects conversational intelligence, voice, and visual performance in a streaming system. Meta also notes that the showcased characters do not all represent avatars available in the app.
Our view is that a face earns its place when the conversation benefits from attention and back-and-forth. Consider a travel agent that has found three possible weekends. You open a call and explain that the cheapest option looks exhausting. It can ask what matters more, walk through the tradeoff, and help you settle on a plan. The itinerary remains visible beside the conversation.
Or imagine a personal agent that helps you prepare for a difficult meeting. You rehearse the opening aloud. It responds, you interrupt, and you try again. A consistent face and voice give the exercise a clear counterpart. The experience is designed around practicing a conversation you will actually have.
A daily check-in offers another starting point. Yesterday you said a project was stuck. Today, the same character opens with a relevant question, using context your application has retained. You can explain what changed without reconstructing the entire story. Familiarity comes from that continuity as much as from the portrait on screen.
These are product scenarios, with outcomes that need testing. A face alone does not establish that users will return more often or finish more tasks. We would look for simpler evidence first: can people correct the agent easily, understand what it is doing, and leave the call with a useful next step?
Build the conversation with Realtime Avatar
Realtime Avatar gives developers the voice and video call experience for their own product. Start with a portrait, create a character, and shape its voice and instructions. The platform supports live conversations and interruptions, so a user can redirect the exchange while the character is speaking.
Your application supplies the personal context. Before a call, it can provide relevant preferences, open tasks, and what the user was trying to accomplish last time. Keeping that context focused helps the character begin with something useful to say. The application decides what to remember across calls and what to leave behind.
The connection to useful work is equally concrete. RTA supports tools implemented in your application and a way for your backend to steer a reply using the result of its own agent logic. A planning assistant might retrieve available times; a project companion might check the next unfinished task. Credentials and action permissions stay with the application that executes those tools.
That lets you build a coherent handoff: the person talks through a goal, your agent does the work, and the character explains the result. Longer jobs can continue in your backend after the call ends. When the person returns, your application brings the relevant progress into the next conversation.
Keep the useful details on screen, too. A spoken explanation can sit beside a calendar, a draft, or a list of choices. Let someone switch to text or voice when that suits the moment. The face should make the product easier to use, and the interface should remain flexible enough for a quick task on a crowded train.
Start with one reason to call
The most promising first version has a specific reason to exist: a morning planning conversation, a rehearsal before a meeting, or a check-in on a goal someone cares about. Give that experience a recognizable character, a useful piece of context, and one action it can help complete.
Instinct, Muse, and Tab show how personal software can grow around an ongoing conversation. For builders, adding a face is an opportunity to make selected moments in that conversation feel more present. Your agent can keep working in the background and still be someone the user knows how to call.
Start in the Realtime Avatar studio, find the character that fits your product, and build one conversation worth coming back to.