Tavus

Today we are introducing PAL Maker, a no-code way for anyone to bring a PAL to life. PALs are a new kind of application- not software you operate with buttons and pages, but something you talk to face-to-face, like a friend or coworker, because the richness of a conversation that carries tone, expression, and timing gets more across, and more done, than any interface before it. A PAL reads the moment, interacts the way we do, and bridges human and machine, whether it is with one person or a thousand. Until now, building a PAL meant a team of engineers, expertise in designing a personality, engineering prompts, and wiring integrations. Even for technical builders, the distance between an idea and a working PAL was real. PAL Maker closes it, for everyone. Building your own PAL works the way describing a person does- with Charlie, the creator of PALs, building alongside you to the same standard our own engineers hold. All you have to do is tell him who they are, how they should sound, and what they are for, and the two of you bring them to life together in the same conversation. Here are the PALs people are already making with Charlie: - A pediatric nurse who had never written a line of code told Charlie, "I want someone warm and unhurried who can answer a family's three-in-the-morning questions after their child's surgery," and by the end of her lunch break the care companion was real, knowing exactly when to bring a human in. - An operations lead described a standup assistant. Now a PAL joins the call on its own, catches the blocker nobody raised, and files the tickets before anyone has closed a laptop. - A teacher described a one-on-one tutor for every kid after school, custom to their level. Now it remembers where each student got lost last week and picks the thread right back up. A PAL is never only a face and a voice. Teaching a machine to be human means teaching it to interact the way we do, in every sense, and underneath sits years of research. It sees and hears you the way a person in the room would, catching the frustration before you have said the word frustrated. It carries the timing of a real conversation, knowing when to speak and when to let silence breathe. It shows you a face that renders not just what it says but how it feels, so you can see it is actually listening. And it remembers what matters about you and lets the rest fall away, because being remembered is most of what it means to be known. All of it built so the machine meets us where we are, with no language of its own to learn first. Whoever you have imagined, a teammate, a tutor, a coach, Charlie is ready to make them. Go ahead and bring your idea to life: https://lnkd.in/g_H7VDqA

Build real-time video agents your users can talk to face to face. See it in a 30-minute demo and pick a free World Cup jersey on the way out.

If your AI could look patients in the eye, what would it say? Build realtime video agents, for any use case, that feel just as human as you do. Built with the world's best perception, rendering, and turn-taking models. Backed by 91,000+ research citations. Have a conversation to see for yourself:

What if your patient app could feel like a real person? Build realtime video agents, for any healthcare use case, that look patients in the eye and answer the question they actually have. Built with the world's best perception, rendering, and turn-taking models. HIPAA compliant. Backed by 91,000+ research citations. Have a conversation to see for yourself:

We didn't build our Conversational Video Interface to be another AI chatbot with a face slapped on top. We built it because conversation isn't text on a delay, it's tone shifting mid-sentence, it's knowing when to pause, it's reading the room before you respond. Most AI agents can't do any of that. They wait for you to stop talking, transcribe what you said, generate a reply, and read it back at you, which is not really a conversation so much as a series of voice memos with extra steps in between. CVI works differently because conversation does. It runs as a closed loop over WebRTC, with seven layers happening at once instead of one after the other. Raven-1 sees and hears you the way a person would. Sparrow-1 handles the rhythm of conversation, picking up turn signals and knowing when to respond and when to wait. Phoenix-4 renders a real face back to you in real time, with the warmth and micro-expressions that make a moment land. All of it happening sub-second, utterance to utterance. The reason this matters is simple. If AI is going to carry the conversations that drive real business outcomes, from sales and support to healthcare intake and onboarding, the way it shows up in those moments can't feel like a transaction. It has to feel like a conversation with someone who is actually there. That's what we're building at Tavus, and it's why we're building the full stack rather than just the face.

If your product could look someone in the eye, what would it say? Build realtime video agents, for any use case, that feel just as human as you do. Built with the world's best perception, rendering, and turn-taking models. Backed by 91,000+ research citations. Have a conversation to see for yourself: