Tavus

Phoenix-4, Raven-1, and Sparrow-1

Tavus builds foundational models that teach machines the art of being human. Phoenix-4 is a gaussian-diffusion based rendering model that synthesizes high-fidelity facial behavior in real time with contextually accurate emotions. Raven-1 is a multi-modal perception model that translates facial expressions, tone, gaze, and emotion into conversational signals. Sparrow-1 is a conversational flow model delivering low-latency turn-taking using lexical, semantic, prosodic, and acoustic cues.

Give AI a face, presence, emotional range, and the ability to see, hear, and understand in real-time.

Build PALs with the API

The Developer API platform brings together building blocks for real-time, interactive PAL conversations: perception, understanding, voice, and rendering. Create and deploy PAL agents that can see, listen, understand, speak, and react naturally. The API includes access to state-of-the-art models, deployment-ready prebuilt components, and a no-code platform.

Developers can integrate human-like AI into any application with full control over the conversation pipeline.

No-code PAL creation

PAL Maker is a no-code way to build and deploy human-like AI. Drop one sentence into PAL Maker or talk with Charlie (the kid inside the machine) and he'll create a PAL for you, ready to deploy in your website or app in just a few clicks.

Anyone can create a custom AI PAL without writing code, making human-like AI accessible to all.

Infinite use cases with infinite scale

A PAL can be a teammate, an onboarding guide, a customer expert, or a trusted companion. Example use cases include L&D agents that pick up where a student got stuck, HR onboarding that welcomes new hires, team meetings that join standups and file tickets, GTM agents that greet visitors and book meetings, customer support that sees problems and ships fixes, and patient intake that hands clinicians a clear picture.

Organizations can deploy PALs across learning, HR, sales, support, and healthcare workflows.

Deploy PALs at scale

Solutions are bespoke, fully managed PAL deployments. Tavus designs, builds, and tunes a PAL around your exact workflow, then runs it in production at the scale and reliability your business requires. All integrations and service guarantees are handled for you.

Enterprises get a custom AI agent tailored to their workflow with full management and support.

Face-to-face video AI

Tavus powers face-to-face video conversations with human-like AI. The platform supports 42+ languages, 1080p resolution, 24kHz audio, and both audio-only and audio+video pipeline modes. It includes features like knowledge base (RAG), memories, objectives and guardrails, function calling, and Google Meet & Zoom integration.

Deliver real-time, emotionally intelligent video conversations that feel natural and human.

• Pioneering Human Computing Since 2020Tavus is an AI research lab pioneering human computing, building foundational models that teach machines to look, see, hear, understand, remember, and respond with emotional intelligence. The result is a new kind of computing that feels instinctive, alive, and deeply human.
• Three Core ModelsTavus offers Phoenix-4, the most natural and realistic human rendering model; Raven-1, a multi-modal perception model for seeing, hearing, and understanding in real-time; and Sparrow-1, a conversational flow model for low-latency, accurate turn-taking.
• PAL Maker and Developer APIPAL Maker is a no-code way to build and deploy a PAL by dropping one sentence or talking with Charlie. The Developer API provides building blocks for real-time, interactive PAL conversations, including perception, understanding, voice, and rendering.
• Use Cases Across IndustriesPALs can serve as team meeting participants, GTM agents, customer support agents, patient intake assistants, L&D agents, and HR onboarding guides, with examples shown for each use case.

• Pricing PlansPricing is available for PALs and Developers. PALs plans include Free ($0/mo), Plus ($20/mo), and Max ($50/mo). Developer plans include Basic (Free), Starter ($59/mo), Growth ($397/mo), and Enterprise (Custom). Specific features and limits are listed for each plan.
• Account TypesUsers can choose between a Developer Account for building with APIs and a PALs Account for personal AI companions. The website states you can switch anytime, but does not detail the switching process.
• Enterprise SolutionsEnterprise solutions are described as bespoke, fully managed PAL deployments where the team designs, builds, and tunes a PAL around your workflow. Integrations and service guarantees are handled, but specific integrations are not listed.
• Model CapabilitiesPhoenix-4 is described as a gaussian-diffusion based model for high-fidelity facial behavior. Raven-1 translates facial expressions, tone, gaze, emotion, and environmental context. Sparrow-1 uses lexical, semantic, prosodic, and acoustic cues for turn-taking.

L&D Agents

A PAL can pick up where a student got stuck last time, walk them through it again, and stay patient for as long as it takes.

HR Onboarding

A PAL can welcome new hires, guide them through company tools and policies, and answer questions from day one.

GTM Agents

A PAL can greet every visitor the moment they arrive, answer what they actually asked, and book the meeting before they click away.

Customer Support

Show a PAL the error light or the cracked part, and it sees the problem, pulls your warranty, and ships the fix before you hang up.

Sign up and upload your brand assets to the central library.

Connect your product feed or CRM via the native integration.

Map data fields to the script template and select an avatar.

Trigger the batch render and review performance metrics in the dashboard.