Tavus
YC S21San Francisco, US Β· Founded 2021 Β· 75 employees Β· Hiring Β· 16 known investors
Tavus develops human-like AI agents called PALs that can see, hear, act, and understand in real-time, offering bespoke, fully-managed deployments for various enterprise workflows.
Also known as Tavus, Inc.
Founders & leadershipΒ· Y Combinator alumni (S21)
Tavus was founded in 2021 by Hassaan Raza and Quinn Favret.
Investors Β· 16
Also in the syndicate Β· 5
Funding
SEC filings, press & company announcements$80M disclosed across 2 of 5 rounds Β· 2023β2025
- $40MSeries BNov 2025 Β· 3 sources
CRV (lead), Flex Capital, HubSpot Ventures, Paul Johnson, Scale Venture Partners, Sequoia Capital, Y Combinator
Source β - $40MraisedNov 2025 Β· 2 sourcesSource β
- Undisclosed amountSeries AMar 2024
Scale Venture Partners (lead), HubSpot, Sequoia, Y Combinator
Source β
Source: company announcements and press reports β follow each round's link for the claim.
Company profile
researched Aug 2026Tavus is a San Francisco-based AI research lab and developer platform working on what it calls "human computing": AI systems that can look, see, hear, remember, and respond with emotional intelligence in real time. Its central abstraction is the PAL (described in one source as "Personal Affective Links"), an AI counterpart that combines behavior, knowledge, memory and pipeline configuration with a rendered "Face" defining on-screen likeness and voice. PALs operate across text, voice and face-to-face video [0][1][5][6].
The company develops its own foundational models in-house: Phoenix-4, a gaussian-diffusion-based real-time human rendering model that generates facial expression, head pose and emotion; Raven-1, a multimodal contextual perception model that interprets faces, gaze, tone, gestures and environment; and Sparrow-1, a transformer-based conversational timing/turn-taking model that uses lexical, semantic, prosodic and acoustic cues. These are paired with an orchestration and memory management system [0][1][5][6].
Products include the Conversational Video Interface (CVI), the developer API that unifies perception, speech, language, rendering and orchestration and supports bring-your-own LLM, retrieval over external knowledge, persistent memory, guardrails and function calling; PAL Maker, a no-code builder for creating and deploying PALs; Faces, photorealistic likenesses rendered by Phoenix-4 from a single image or roughly two minutes of video, plus a stock face library; and Solutions, enterprise deployments built and managed by the Tavus team. Documentation notes the API has moved to "PAL" and "Face" terminology while retaining legacy "persona" and "replica" endpoints for backward compatibility [0][1][3][6]. Stated platform characteristics include sub-600ms utterance-to-utterance latency, support for 43 languages, and SOC 2 Type 2, HIPAA and GDPR compliance [6].
Founding story
Sources conflict on the founding year: Y Combinator and aggregator profiles list 2021 (YC Summer 2021 batch) [2][4], while Tavus's own information page and UpMarket state the company was founded in 2020 by Hassaan Raza and Quinn Favret, with the website noting it has been "pioneering human computing since 2020" [0][1][6][7]. The stated motivation was that automation had scaled efficiency while removing empathy, nuance and presence from digital interactions; the founders' YC launch post framed the problem as consumer isolation amid saturated digital products and services [2][6].
Business model
Tavus sells developer access to its models through APIs (the Conversational Video Interface) and a no-code builder (PAL Maker), alongside a managed "Solutions" offering in which the Tavus team builds, tunes and operates enterprise deployments for use cases such as sales, healthcare, learning and development, and recruiting. It also offers consumer-facing PAL accounts for individuals in addition to developer accounts [0][1][6].
The company lists Free, Starter, Growth and Enterprise pricing tiers, with included conversation minutes and faces and discounts at scale; enterprise deployments are delivered as managed engagements [6].
Traction
Tavus states it powers more than 100,000 developers [0][1][6] and lists enterprise users including Amazon, Deloitte, EY, Mayo Clinic, CVS, Salesforce, Aetna and Wix [6]. Published customer examples include Final Round AI (mock interviews for 100K+ users and 1.2M practice minutes on CVI), Qualified (its Piper AI SDR, reporting 2x more conversations and 6x more meetings booked), Orum (sales role-play), ACTO (life-sciences rep training), Delphi (live face-to-face video calls) and CareFlick (senior companionship) [6]. Y Combinator lists a team size of 75 [2].
Latest developments
In November 2025 Tavus announced a $40M Series B led by CRV with participation from Scale Venture Partners, Sequoia Capital, Y Combinator, HubSpot Ventures and Flex Capital, to be used to expand operations and development efforts [5][4][6]. The company also launched PAL Maker, a no-code way to build and deploy human-like AI, and updated its API terminology from persona/replica to PAL/Face [0][1][3][6].
βΈFull profile β market position, technology, go-to-market, geography, history, risks & controversies
Market position
Tavus positions itself as a leader in human simulation models and distinguishes its real-time conversational stack from asynchronous AI video generation tools, emphasizing that its pipeline is built on its own foundational models rather than assembled from third-party components [6]. One aggregator profile describes it as building the "human layer" of AI with lifelike digital twin and real-time conversational video experiences [4][7].
Cited differentiators are real-time realism (sub-600ms latency with emotionally expressive rendering), an end-to-end stack spanning perception, memory, knowledge and action built on proprietary models, flexibility via bring-your-own LLM and modular APIs plus a no-code builder, scalability from individual to enterprise deployments, and compliance/consent mechanisms (SOC 2 Type 2, HIPAA, GDPR) [6]. Its YC launch post claimed the Conversational Video Interface was the only end-to-end conversational pipeline with under one second of latency [2].
Technology
Proprietary foundational models built in-house: Phoenix-4 for real-time full-duplex human rendering with emotional expression and identity preservation; Raven-1 for multimodal perception fusing tone, prosody, expression, posture and gaze; and Sparrow-1 for conversational timing and interruption handling within tens of milliseconds. These run behind the Conversational Video Interface pipeline with orchestration, memory, retrieval, guardrails and function calling, delivering sub-600ms utterance-to-utterance latency and 43-language support [0][1][5][6].
Go-to-market
Distribution is developer-led β self-serve developer accounts, public APIs, documentation and demo agents (e.g. Charlie and, earlier, Carter), plus a Discord community and live demos β combined with a direct enterprise motion through Tavus Solutions for managed deployments. A separate consumer PAL account track targets individuals [0][1][2][3][6].
Product teams and developers embedding conversational video into applications; enterprise teams in sales, healthcare, learning and development, recruiting and customer support; and individuals or prosumers using PALs as tutors, assistants or companions. Named industries include healthcare, recruiting, sales, education and customer engagement [2][4][6].
Geography
Headquartered in San Francisco, California, United States, with roles posted in San Francisco and remote within the US [2][4][6][7]. The platform supports 43 languages [6].
History
Tavus went through Y Combinator's Summer 2021 batch [2]. Press coverage tracked by YC includes a March 2023 TechCrunch piece on generative-AI-powered personalized videos with voice and face cloning, an $18M raise reported in March 2024, and a $40M raise reported in November 2025 [2]. The company launched Conversational Replicas / the Conversational Video Interface, enabling live video conversations with digital twins that speak, see and hear [2][4]. It subsequently rebranded its category around "human computing," renamed personas and replicas to PALs and Faces in its API and documentation, released the Phoenix-4, Raven-1 and Sparrow-1 models, and shipped PAL Maker, a no-code builder [0][1][3][5][6].
Risks & controversies
The sources do not report controversies. Tavus notes it builds disclosure and consent mechanisms into products handling synthetic likenesses and cites SOC 2 Type 2, HIPAA and GDPR compliance [6]. Secondary-market listings note Tavus shares are illiquid, speculative and carry risk of total loss, with no public market or guaranteed exit [7]. Public sources also disagree on the founding year (2020 vs. 2021) and on total funding disclosed [2][4][6][7].
Compiled by commissioned research from 8 cited public sources β announcements, filings, and press listed under research sources below.
Key figures
latest reportedCompany-reported or press-reported figures, each dated to when it was claimed β not independently audited.
Competitors Β· 4
by search overlap
Companies competing with Tavus for the same Google search keywords, organic and paid, via search-intersection analysis.
Timeline Β· 8
launches, deals, and filingsTavus documentation states the API and docs now use "PAL" for behavior, knowledge and pipeline configuration and "Face" for visual appearance and voice, with legacy persona and replica endpoints and fields retained for backward compatibility.
PAL Maker, a no-code builder for creating and deploying human-like AI PALs, went live; users describe an idea to a built-in guide named Charlie, assign a face, voice, knowledge and capabilities, then deploy to a site or app.
San Francisco-based human computing company raised $40M in Series B funding led by CRV, with participation from Scale Venture Partners, Sequoia Capital, Y Combinator, HubSpot Ventures and Flex Capital. Proceeds are earmarked for expanding operations and development efforts.
$40M source β
Tavus described its in-house model suite powering PALs: Phoenix-4 for real-time rendering with expression, head pose and emotion control; Sparrow-1 for audio and semantic conversational timing; and Raven-1 for contextual perception of people, environments, emotions and gestures.
Tavus launched the Conversational Video Interface, giving developers tools to build live video conversations with digital twins that speak, see and hear, with under one second of latency; demo digital twin "Carter" was made available on the website.
Series A round; TechCrunch reported Tavus confirmed an $18M Series A led by Scale Venture Partners and said it was opening its platform for third-party integrations. Y Combinator's news list dates the coverage of this raise to 2024-03-12.
$18M source β
Seed round with participation from Scale Venture Partners, Accel, Index Ventures, Lightspeed Venture Partners, YC Continuity and angel investor Priyank Chandra.
$6M source β
Tavus is listed as a Y Combinator Summer 2021 company with primary partner Diana Hu.
Dated company events from announcements, filings, and press; legal rows summarize public dockets and regulator releases.
In the news
βΈResearch sources Β· 8
primary sources listed
- Tavustavus.io Β· web
8 public sources were cited for this profile; the first-party ones are listed here.
Frequently asked questions
- What does Tavus do?
- Tavus builds foundational AI models and APIs for real-time, face-to-face conversational AI humans it calls PALs.
- Who founded Tavus?
- Tavus was founded by Hassaan Raza, Quinn Favret in 2021.
- Who are Tavus's investors?
- Tavus's investors include CRV (Charles River Ventures), High Alpha, HubSpot Ventures, Mantis Capital, Scale Venture Partners, Sequoia Capital, Seven Stars Ventures, Y Combinator and 3 more.
- How much funding has Tavus raised?
- Tavus has disclosed $80M raised across 2 of its 5 known rounds.
- Where is Tavus headquartered?
- Tavus is headquartered in San Francisco, US.