What Tavus does

Tavus is a privately held AI research lab founded in 2020 and headquartered in San Francisco that builds emotionally intelligent, real-time conversational AI humans for enterprise, developer, and consumer use cases. Its core technology stack comprises three proprietary foundational models — Phoenix-4 (a hybrid Gaussian-diffusion real-time human rendering model running at 40fps in 1080p with 10+ controllable emotional states), Raven-1 (a multimodal perception model fusing audio and visual signals at sub-100ms latency), and Sparrow-1 (a conversational turn-taking model with 55ms median floor-ownership latency) — packaged into the Conversational Video Interface (CVI) developer platform and the PALs (Personal Affective Links) consumer AI companion product. The company serves enterprise customers across learning & development, healthcare, sales, education, and support workflows; developers and product teams integrating the APIs; and individual consumers seeking persistent AI companions.

Tavus monetizes through four revenue streams: usage-based pricing for the developer API and CVI platform; enterprise subscriptions for production-grade PALs and video-agent deployments with custom replicas and SLAs; recurring PALs consumer subscriptions; and a freemium onboarding funnel (free Developer and PALs accounts) feeding paid upgrades. Its go-to-market blends product-led growth via self-serve developer sign-up with a demo-led enterprise sales motion, supplemented by channel partnerships including Ericsson's telecom Corina avatar assistant and infrastructure partnerships with Cerebras Systems for low-latency inference.

The company is venture-backed, having raised $65 million cumulatively through a November 2025 $40 million Series B led by CRV, with participation from Sequoia Capital, Scale Venture Partners, and Y Combinator, alongside prior rounds led by Sequoia and Scale. Its cap table reflects a syndicate of tier-1 institutional investors, and the company claims to power 100,000+ developers while displaying enterprise reference logos including Deloitte, Amazon, Salesforce, CVS Health, and OpenAI's ChatGPT. Headcount sits at 51-100 employees with active hiring, and the company maintains compliance posture across GDPR (EU/UK), biometric data, and WCAG 2.2 Level AA accessibility remediation.

Name

Tavus

Legal name

Tavus Inc.

Website

https://tavus.io

Company type

Private

Founded year

2020

Operating status

Operating

Headcount range

51–100 employees

Short description

Tavus is a San Francisco-based AI research lab, founded in 2020, that builds proprietary foundational models for emotionally intelligent, real-time conversational AI humans, serving enterprise customers across healthcare, learning, sales, and support, plus developers via its CVI API platform and consumers through its PALs companion product.

Ownership category

akta.pro rank

Product category

Conversational Video AI Platform

NAICS

Custom Computer Programming Services (541511), Software Publishers (5132)

SIC

Services-Computer Programming Services (7371), Services-Computer Integrated Systems Design (7373)

akta.pro primary industry

Conversational AI Platforms (chat/voice bots, orchestration) (HDAAAFAC)

akta.pro secondary industries

AI Application Enablement Platforms (Copilot/Agent Frameworks, SDKs) (HDAEANAJ), Voice Assistants & Voice UI (VUI) (HDAAAFAD), Video Conferencing & Meetings (HDAIAHAI)

Keywords

• Conversational video AI
• AI digital humans
• Developer video API
• Multimodal AI models
• AI avatar platform

Headquarters

HQ city

San Francisco

HQ country

United States

HQ region

North America

Offices2 records

Markets served

GTM type

B2B and B2C

Offering type

Software

Cost components

Technology or R&D, Personnel, Infrastructure, Marketing or Sales, Operations

Revenue model

1. PALs consumer subscriptions: Recurring subscription access to Tavus's PALs platform for individuals ("Best for individuals looking to talk, explore, and connect with a friend"), with a dedicated PALs pricing page and a "Get Access" entry point.
2. Developer API and Conversational Video Interface (CVI) usage: Usage-based revenue from developers and enterprises integrating Tavus APIs, SDKs, and the CVI platform (video generation, conversational video, perception and rendering models) into their own products. Self-serve developer sign-up with a free tier and paid plans available via platform.tavus.io.
3. Enterprise video agent and PALs deployments: Production-grade enterprise deployments of Tavus's video agents and PALs with custom replicas, emotion control, and enterprise SLAs, sold via demo-led enterprise sales motion across learning & development, healthcare, sales, education, and support.
4. Free tier / freemium onboarding: Free Developer Account and free PALs Account tiers available at sign-up ("Get Started" with plan=free), used as a freemium funnel into paid usage.

Pricing tiers

Go-to-market motion1 record

Distribution channels5 records

Marketing channels10 records

Core offering

Tavus builds and sells a real-time conversational video AI platform composed of three proprietary foundational models (Phoenix-4 rendering, Raven-1 perception, Sparrow-1 turn-taking) that power two go-to-market offerings: a developer-facing Conversational Video Interface (CVI) with APIs, SDKs, and a Video Generation API for embedding AI humans into third-party products, and a consumer-facing PALs (Personal Affective Links) companion product with multimodal chat, voice, and video. Enterprise customers additionally license production-grade video agents across learning & development, healthcare, sales, education, and support workflows.

Product overview

Tavus operates a platform-plus-modules architecture centered on its core CVI (Conversational Video Interface) developer platform and the consumer-facing PALs (Personal Affective Links) product. The CVI developer platform exposes Tavus's proprietary AI models — Phoenix-4 for real-time emotional rendering, Raven-1 for multimodal perception, and Sparrow-1 for conversational flow — as APIs and SDKs, alongside a Video Generation API and a Stock Replica library, letting developers build and deploy custom interactive avatars. PALs is Tavus's finished consumer AI companion product built on the same underlying models, offering ready-to-use multimodal communication, memory, agentic capabilities, and Google Workspace integrations. Enterprise solutions include the AI SDR sales automation avatar and dedicated vertical offerings (healthcare, sales, education, support, L&D). Tavus also operates limited-time consumer experiences like AI Santa 2.0, delivered through the PALs platform.

Differentiator

Problem solved

Functional benefit

Brands

• PALs (Personal Affective Links): Emotionally intelligent AI humans that see, listen, remember, and grow; supports multimodal communication (video, voice, text chat, SMS), 30+ languages, memory, and integrations with Google apps — sold as a consumer-facing companion product.
• Conversational Video Interface (CVI)

Products and services

• PALs (Personal Affective Links) Consumer-facing emotionally intelligent AI companion product that sees, listens, remembers, and grows. PALs offer full multimodal communication (video, voice, text chat, SMS) with 200–600ms response times, interruptibility, perception, memory, evolving relationships, proactivity, support for 30+ languages, agentic capabilities (taking actions, async tasks, planning, multi-step execution), and Google Workspace integrations. Sold to individual end-users and enterprises via self-serve sign-up at platform.tavus.io and a dedicated pricing page at tavus.io/pals/pricing.
• Conversational Video Interface (CVI) Developer-facing platform that exposes Tavus's proprietary Sparrow-1 (turn-taking), Raven-1 (multimodal perception), and Phoenix-4 (real-time rendering) models as APIs, SDKs, and guardrails so developers can build and deploy conversational video agents that see, listen, understand, speak, and react naturally. Includes replicas, personas, LLM hooks, TTS engines, perception layers, knowledge sources, and custom function/tool calls.
• Phoenix-4 Real-time human rendering model producing emotionally responsive, context-aware human presence in live conversations. The first model to generate and control emotional states, active listening behavior, and continuous facial motion as a unified system. Runs at 40fps in 1080p on a hybrid Gaussian-diffusion architecture with an implicit motion space enabling emergent micro-expressions. Available through the Tavus platform, APIs, PALs, and the Stock Replica library.
• Raven-1 Multimodal perception model giving machines the ability to see, hear, and understand in real-time. Fuses audio and visual signals into natural-language descriptions of user state, intent, and context at sentence-level granularity with sub-100ms latency, keeping rolling perception no more than 300ms stale.
• Sparrow-1 Conversational flow model delivering the lowest-latency and most accurate turn-taking. Predicts who owns the conversational floor at every moment using lexical, semantic, prosodic, and acoustic cues such as pauses, filler words, intonation, and trailing thoughts. Median floor-ownership latency of 55ms with dynamic response latency under 100ms when confident, typically 200–500ms.
• Video Generation API Developer API for generating AI-personalized videos at scale as part of the Tavus developer platform, enabling automated personalized video creation and distribution for sales, marketing, and investor outreach.
• AI SDR Enterprise AI Sales Development Representative solution for automating personalized outbound sales outreach via Tavus's video AI capabilities, enabling human-like AI agents to perform 1:1 sales messaging at scale.

Quantifiable outcome

• 200-600ms response times for PALs conversations with sub-100ms speculative-inference capability when confident
• +7 more outcomes

Named customers9 records

Segments4 records

Ideal customer profiles4 records

Technology focussed Yes

API detail

Has API

Yes

API docs

API detail

Core technology

AI maturity

App detail

Integration4 records

AI capability13 records

Feature6 records

Partnerships

Eleven partnerships are on record, tiered flagship enterprise / telecom partnership, co-deployed partner via ericsson corina, enterprise security partner, native platform integration, core inference infrastructure partner, operational vendor (payment processing), operational vendor (subscription billing), operational vendor (session analytics) and operational / compliance vendor.

• Ericssonflagship enterprise / telecom partnershipStrategic or Co-development Partner · 23 June 2026Ericsson powers its enterprise avatar assistant "Corina" for telecom customer service through a partnership with Tavus (avatar technology), Cartesia (text-to-speech, an Ericsson investment), and Pipecat (virtual room platform). Demoed live connecting Sweden and Canada, targeting telecom operator revenue and churn reduction.
• Cartesiaco-deployed partner via ericsson corinaTechnology or Integration · 23 June 2026Provides text-to-speech for Ericsson's Corina enterprise avatar solution alongside Tavus avatar technology. Cartesia is an Ericsson investment.
• Pipecatco-deployed partner via ericsson corinaTechnology or Integration · 23 June 2026Provides the virtual room platform underlying Ericsson's Corina enterprise avatar solution, connecting Tavus avatars and Cartesia voice in live demos.
• SecurityPal AIenterprise security partnerTechnology or Integration · 19 February 2026Tavus adopted SecurityPal AI's Assurance Management Platform to handle enterprise security questionnaires and biometric data requirements when selling to large enterprises, turning security review into a competitive differentiator.
• Google Workspace (Gmail, Calendar, Drive, Docs, Sheets)native platform integrationTechnology or Integration · 12 November 2025Connected-services integration via Google's OAuth 2.0 process and Google's Limited Use Policy, enabling PALs to summarize emails, update documents, and manage events at the user's request. Conversation logs may include connected-account data and are retained up to 90 days.
• Cerebras Systemscore inference infrastructure partnerTechnology or Integration · 7 March 2025Cerebras's CS-3 Wafer Scale Engine 3 inference platform (2,100 tokens/second on Llama 3.2 70B) powers Tavus's real-time AI applications, enabling significant latency reductions alongside customers GSK, Audivi, Vellum, and LiveKit.
• Stripeoperational vendor (payment processing)Technology or IntegrationStripe processes payment card information for Tavus purchases directly; Stripe may use payment data per its privacy policy (https://stripe.com/privacy).
• Recurlyoperational vendor (subscription billing)Technology or IntegrationRecurly processes payment data for Tavus subscriptions per its privacy policy (https://recurly.com/legal/privacy).
• Hotjaroperational vendor (session analytics)Technology or IntegrationProvides session-replay technology used by Tavus to record user interactions on its websites for product improvement; opt-out available via Hotjar's policies page.
• LogRocketoperational vendor (session analytics)Technology or IntegrationProvides session-replay technology alongside Hotjar for analyzing user behavior on Tavus's Service.
• DataRep (EU & UK GDPR representatives)operational / compliance vendorOthersDataRep serves as Tavus's appointed representative in the EU and the UK for GDPR purposes; users in Europe can contact DataRep for Tavus data-processing queries.

Scale indicators11 records

Recent moves6 records

Expansion highlights6 records

Direct peers

• HeyGen: HeyGen is a direct competitor offering AI-generated avatar videos and conversational video agents for enterprise use cases (marketing, sales, training). It targets the same developer and enterprise buyer with similar generative avatar technology and overlapping stock-replica and custom-avatar capabilities.
• Synthesia: Synthesia is a direct competitor in AI-generated talking-head video for enterprise L&D, sales, and marketing. It shares Tavus's enterprise GTM motion, custom avatar offering, and multilingual value proposition, though it is more established in pre-rendered rather than real-time conversational video.
• D-ID: D-ID builds AI-powered talking avatars and conversational video interfaces for enterprise and developer customers, overlapping directly with Tavus's CVI platform and stock-replica library. It competes on similar talking-head rendering and digital-twin capabilities.
• Soul Machines: Soul Machines builds emotionally intelligent digital humans for enterprise customer experience, healthcare, and education — the same emotionally-aware avatar niche as Tavus. Both companies compete on behavioral realism and trust-driven enterprise deployments.
• ElevenLabs: ElevenLabs is a leading AI voice and conversational AI platform whose voice and speech capabilities overlap with Tavus's audio pipeline (Sparrow-1 turn-taking, TTS integrations). It is a direct competitor for voice-layer enterprise and developer budgets and increasingly a multimodal conversational AI peer.

Emerging players

• Hume AI: Hume AI builds emotionally intelligent voice and expression models with a focus on measuring and responding to user affect. It overlaps with Raven-1's multimodal perception and Sparrow-1's prosody-aware turn-taking, and is an emerging peer in the emotional-AI category.
• Character.AI: Character.AI operates a consumer-facing conversational AI companion platform that overlaps with Tavus's PALs consumer product in target audience and conversational memory features. It is an emerging peer in the consumer AI-companion category rather than in the enterprise video-agent category.
• Replika: Replika is a consumer AI companion app with persistent memory and emotional engagement, directly comparable to Tavus's PALs product. It is an emerging peer in the consumer AI-companion space and provides a comparable demand-signal benchmark.

Broad incumbents

• OpenAI: OpenAI is a broad incumbent offering multimodal conversational AI APIs (GPT-4o, the Realtime API) that compete with Tavus's CVI at the conversational AI orchestration layer. Listed on Tavus's homepage as a 'partner/customer,' it is simultaneously a potential partner and the largest competitive threat to Tavus's developer API.
• NVIDIA (Avatar Cloud Engine / ACE): NVIDIA's ACE and Tokkio offerings provide cloud-native digital-human and avatar infrastructure for enterprises and telcos, competing with Tavus at the platform layer for telecom and enterprise deployments. NVIDIA is a broad incumbent whose moves in conversational avatars directly affect Tavus's addressable market.

Market position

Strengths5 records

Weaknesses5 records

Competitive moat5 records

Key risks6 records

Key highlights7 records

Customer concentration

• Xx.com/heytavus
• YouTubeyoutube.com/channel/UCMDJzuCOW22MKAaHXwMzvGA
• LinkedInlinkedin.com/company/tavus-io

Compliance3 records

Revenue estimate

Valuation estimate

Number of profiles

Profiles4 records

Subsidiaries2 records

Funding overview

Funding rounds5 records

Investors25 records

Funding detail is available on the Subscription and Enterprise plan.Contact sales →

M&A

Investments

M&A and investment is available on the Subscription and Enterprise plan.Contact sales →

Frequently asked questions about Tavus