Tavus

Tavus is building real-time AI humans — systems that can see you, hear you, and respond with natural expression, emotion, and context. What began as personalized video has grown into a full platform used by companies from startups to the Fortune 10. The team recently raised a $40M Series B to advance this vision, introducing PALs: agentic AI humans that can perceive, reason, and act on their own. In this conversation with YC’s Diana Hu, founders Hassaan Raza and Quinn Favret share how they made the leap from generative video to real-time AI humans, the foundational models behind rendering and perception, and why they believe AI humans will become the next major interface for work and communication. Learn more about Tavus at https://www.tavus.io. Chapters: 00:24 – From Personalized Video to AI Humans 01:18 – Why Real-Time Matters 02:36 – How AI Humans See, Hear, and Respond 04:05 – Introducing PALs: Agentic AI Humans 05:42 – The Foundational Models Behind Tavus 07:28 – Building Emotion, Expression, and Context 09:10 – Use Cases From Startups to the Fortune 10 11:00 – Raising the $40M Series B 12:52 – The Future: AI Humans as the Next Interface

Introducing Phoenix-4.5, the world's fastest, most realistic AI human. PALs now move with the words they say. Phoenix-4 made the face feel alive. Phoenix-4.5 gives the person room to move. It is the fastest and most expressive real-time human rendering model on the market: 134 ms from audio to video, and for the first time it renders the full upper body instead of the head alone, so expression runs through the head, shoulders, and posture, lip sync is measurably tighter, and emotion carries through listening as much as speaking. The stiffness between the words, the thing that makes other models feel off, is gone. It is the closest AI has ever come to passing the Turing test face to face. WHAT'S NEW - From face to person. A new generative renderer, built from scratch, generates the whole frame in a single pass: face, head, shoulders, torso, and background together. No mask, no seam, no face boundary for motion to expose - More natural, by a wide margin. Richer facial animation, more expressive emotion and micro-movements, tighter lip sync, and listening that looks like listening - Our first zero-shot Phoenix model. A single image or video is enough to create a new face. The preview is usable in about a minute, and fine-tuning finishes in about two hours, half the time Phoenix-4 needed - Far more people can become PALs. Long hair in front of the shoulders, glasses, earrings, headbands, and heavier beards used to fail training. Now they work - Everything that made Phoenix-4 one of a kind carries over: real-time emotion control, active listening, and seamless transitions between listening and speaking, now rendered through the upper body BY THE NUMBERS - 134 ms from audio to video, 25% faster than any other real-time human rendering model - Lip sync up 16% over Phoenix-4 - 1.3x more head and upper-body motion, with identity fidelity held level with Phoenix-4 - 64% of the faces that failed on Phoenix-4 now train successfully - Human preference: zero-shot Phoenix-4.5 scores 1002 Elo, ahead of fine-tuned Phoenix-4 at 929. Fine-tuned Phoenix-4.5 climbs to 1069 WHAT IT UNLOCKS Presence is part of the product. A tutor holds attention through a full lesson. A health guide visibly listens while someone explains something difficult. A sales PAL leans into the pitch and reacts to the objection while it's still being spoken. Builders on Tavus who replaced audio-only interactions with face-to-face AI are already seeing 2 to 3x higher engagement across sales and healthcare, 40% higher knowledge retention in L&D, and 50% faster ramp with 3x higher rep conversion in sales training. Phoenix-4.5 is live today for every PAL through PAL Maker and the API, alongside a new generation of stock faces. Every existing stock face has been upgraded too. TRY IT 🔗 Build a PAL with Phoenix-4.5: https://maker.tavus.io 📖 Read the research from Minh Anh Nguyễn and Haiyao Xaio: https://www.tavus.io/blog/phoenix-4-5 📚 Docs: https://docs.tavus.io ABOUT TAVUS Tavus is the human computing company. We build the models and platform behind PALs: AI humans that see, hear, and converse naturally in real time. For decades, people have adapted to machines. We're teaching machines how humans work. #ConversationalAI #VideoAI #VoiceAI #Tavus #Phoenix45 #PALs