Realtime Conversational Video with Pipecat and Tavus — Chad Bailey and Brian Johnson, Daily & Tavus
AI Engineer World's Fair 2025 · 18:46
Conversational AI and real-time video agents
Tavus builds AI agents called PALs that interact through chat, voice and real-time face-to-face video. Developers use its Conversational Video Interface (CVI) to embed these interactions in products, connecting their own language models, retrieval, memory and tools. PAL Maker lets users create agents without code; Deployments publishes them as website widgets, embeds or hosted links. Tavus also builds and manages enterprise deployments for sales, healthcare, training and recruiting, supporting workflows such as patient intake, customer onboarding and interview practice.
Founded in 2020 by CEO Hassaan Raza and COO Quinn Favret, Tavus develops models for the visual and conversational behavior behind these agents. Phoenix generates facial expressions and movement in real time using Gaussian-diffusion rendering, while Raven interprets cues including tone, gaze and expression. Its Sparrow research addresses conversational timing: Sparrow-2 jointly models speech meaning, speaker identity, interruptions and surrounding audio to decide when to listen or speak, preserving environmental context rather than reducing the task to silence detection.
In 2026, the company reported supporting more than 100,000 developers. Tavus raised a $40 million Series B led by CRV in 2025, bringing total funding to approximately $64 million.
AI Engineer World's Fair 2025 · 18:46
Affiliations reflect their AIE appearances, not necessarily current employment.
The joint discussion covers Pipecat's frames, processors, and voice activity detection alongside transport options and audiovisual synchronization, detailing the infrastructure behind a conversational video agent.
Affiliations reflect each recorded session, not necessarily current employment. The session is a joint Daily–Tavus discussion.