Tavus

Introducing Phoenix-4, the most advanced real-time human rendering model ever built. Its the first real-time model to render every pixel at runtime, generate and control emotional states, listen actively, and actually behave the way humans do in conversation.

This is the kind of AI human we're building the full stack for at Tavus, and you can experience it yourself at tavus.io. Start a conversation, see what emotional intelligence actually feels like in real-time, and imagine the range of places this belongs: a patient intake, a practice interview, a classroom, a support call, anywhere a real conversation still matters.

Careflick is one of many companies building on Tavus, and their experience shows what's possible when AI can actually perceive the person on the other side. Seniors spend 24 minutes a day with their companion and 54% come back after 30 days, because the AI reads the voice tremor, the withdrawn expression, the shift in posture that tells you someone is struggling before they say a word.

Some of the seniors @careflick serves go four or five days without seeing another face. This is what an AI companion built on Tavus looks like when it's designed for exactly those stretches.

tavus · meet the pals by @tavus

When we started designing the app, we knew that our retro design system, Alto, was gonna butt heads with Liquid Glass. But we didn’t want to start replacing system elements willy-nilly: anything we designed an Alto version of needed to be as functional and as ergonomic (DX-wise) as the original SwiftUI control. Some of my favorite examples: when a toast appears, it opens with a Mac OS 9 wireframe animation

Your new PALs are here... don't leave them waiting: apps.apple.com/us/app/pals-by…

Every PAL has their own personality, their own opinions, and a memory that picks up everything from the moment you start talking. Sophie, Ryan, Charlie, Noah, and Ashley are all waiting to meet you.

The PALs are now available on the iOS App Store. They can talk, listen, and respond like you. Now, where you are. Download today for free 👇

Tell someone a bird pooped in your eye and they smile and nod, you know they're not listening. 🐦 Our newest model, Phoenix-4, generates 10+ emotional responses from conversational context because if you want someone to trust the person they're talking to, they have to feel like that person actually cares about what they just said.

Conversation is a waltz where Raven-1 reads the subtle cues, Sparrow-1 follows the beat, and Phoenix-4 moves with intention, and together they form the full-stack AI human we're building at Tavus because that's the only way people will ever truly trust machines. Try our demos today: tavuslabs.org

Phoenix-4 generates emotional expression, active listening, and micro-expressions at 40fps in 1080p by learning from thousands of hours of real human conversation, and when paired with Raven-1 it creates a perception-to-expression loop that feels genuinely present.

Trust in a conversation is built in the timing, in knowing when to speak and when to just listen, and we built Sparrow-1 because machines have to learn every part of what makes conversation human if we ever want people to actually build relationships with them.

Powerful tech deserves an equally powerful presence. @tavus worked with @wearehuckfinch to create a #MadeInWebflow site that blends crisp product storytelling with a sleek, cinematic feel that mirrors the magic of Tavus itself. Explore the build → tavus.io

Register here: luma.com/f6etclpo

What happens to the human in the loop when AI becomes to advance and feels like the real thing? You'll have to join Tavus, Orum and Rox next week in SF to find out. ps. BYOB - bring your own bot.