Skip to main content

Tavus

Freemium

AI that looks and feels human.

Tavus is a San Francisco-based AI research lab pioneering human computing. It builds foundational models that let AI see, hear, and respond like people do, turning the promise of science fiction into reality where computing feels instinctive and truly alive. The platform powers two core offerings: PALs, emotionally intelligent AI companions for consumers, and CVI, a developer-first API platform for real-time conversational video agents.

conversational videoreal timedigital avatarapimultimodalVideo GenerationDigital Avatar GenerationConversational Q&ASpeech RecognitionTask ExecutiongenerationrecognitioneditingVideo CreationAI AgentsAudio & VoiceWorkflow Automation
Visit Website
Tavus preview

What is it

Tavus is a human computing company that teaches machines the art of being human. Its frontier models and APIs enable the creation of digital twins that power advanced video generation and real-time video conversations. The company is backed by Sequoia, Scale VC, and Y Combinator.

What it can do

Tavus offers two distinct products. PALs are personal AI companions that listen, remember, and are always present across text, voice, and face-to-face video. CVI (Conversational Video Interface) is an API-first platform that lets developers embed real-time, human-like AI video conversations into their own applications, complete with multimodal perception, persistent memory, knowledge bases, and function calling.

Who is it for

PALs are designed for individuals looking to talk, explore, and connect with a friend. CVI is built for developers, founders, and enterprise teams who want to integrate real-time AI video into products for healthcare, recruiting, customer support, training, and insurance use cases.

Key Features

Real-Time Conversational Video – Face-to-face AI that sees and hears

CVI delivers real-time multimodal video conversations where AI agents perceive facial expressions, tone of voice, and environmental context through Raven-1, respond with human-like turn-taking through Sparrow-1, and render photorealistic facial behavior through Phoenix-4 at 40fps and 1080p. The full pipeline runs from perception to rendering in under a second, producing agents that feel present and emotionally responsive rather than robotic.

Digital Replica Creation – Train your own AI twin in minutes

Create a custom digital replica from just two minutes of recorded video. The system captures subtle details like how your mouth forms specific phonemes and how your eyebrows move when emphasizing points. Alternatively, choose from over 100 high-quality stock replicas for quick prototyping and testing without any training overhead.

Multimodal Perception – Raven-1 reads the room in real time

Raven-1 is a novel multimodal perception model that unifies object recognition, emotion detection, gaze tracking, and adaptive attention within a single contextual framework. It fuses visual input, emotional signals, and spatial relationships to understand not just what users say, but how they say it and what their environment reveals.

Conversational Flow Engine – Sparrow-1 manages natural dialogue

Sparrow-1 is a transformer-based dialogue model that captures conversational timing, responsiveness, and humanlike interaction flow. It achieves 55ms median floor-prediction latency with zero interruptions, enabling natural turn-taking, intelligent interruptions, and active listening cues that mirror real human conversation rhythm.

Use Cases

Healthcare Patient Intake

Deploy an AI persona for initial patient intake, post-visit education, medication guidance, and appointment preparation. The agent perceives patient expressions to detect confusion or distress, retrieves relevant medical guidelines from the knowledge base, and escalates to clinicians when judgment is required.

Recruiting and Screening

Use a consistent AI interviewer for candidate screening, role explanation, and culture introduction. The agent asks role-specific questions, reads candidate responses and body language, and generates structured assessment data regardless of where the candidate sits or what time zone they are in.

Insurance Claims and Policy Support

Handle claims explanations, policy questions, and renewal discussions with an AI persona that stays within predefined guardrails. Objectives keep conversations purposeful while compliance boundaries prevent the agent from giving regulated advice outside its scope.

Sales Training Roleplay

Create realistic AI roleplay partners that mirror your best managers. Trainees practice objection handling, discovery calls, and closing techniques with an agent that adapts its style, provides real-time feedback on tone and wording, and remembers each trainee's recurring weaknesses across sessions.

Pricing plans

Frequently Asked Questions

Discussion

No comments yet. Be the first to start the thread.