Glossary of common terms

Category (face-to-face, agentic)

Human computing: The movement to close the gap between people and machines by making technology feel unmistakably human. Tavus is “a research lab pioneering human computing.”

AI human: The primary object in our ecosystem: a lifelike, face‑to‑face interface that sees, hears, understands, and responds in real time. Use “AI human” as the default external term; it’s our human‑centric alternative to generic “agents/bots.”  

Teaching machines to be human: Our point of view in five words—appears in short blurbs and company descriptors.

Humanlike presence: What Tavus gives your product: lifelike presence that deepens engagement, trust, and feedback—“AI humans for everything… scale emotional intelligence.”

Products

Conversational Video Interface (CVI): An interface that brings face‑to‑face conversation into any product. CVI doesn’t just stream video; it sees, hears, and understands in real time to drive emotionally intelligent interactions (sub‑second latency). Ideal for teams embedding this natively in their apps.  

Replica (Personal / Stock): The rendered on‑screen identity an AI human uses. “Personal” = trained from a creator’s footage; “Stock” = pre‑built library of optimized identities. Fully white‑labeled; fast training; identity preservation.

AI Persona (Persona): A configured role and behavior profile for an AI human—personality, objectives, guardrails, knowledge, memories. Built via Persona Builder or API.

Core models (the “Tavus brain”)

Phoenix‑4 (face rendering): Full‑face, real‑time generation with pixel‑level identity preservation, studio‑grade lip‑sync, and micro‑expressions for true realism. Powers Replicas in CVI and video generation.

Raven‑1 (perception & vision): Contextual visual understanding in real time: interprets emotion in natural language, maintains ambient awareness, detects specified gestures/objects, and can trigger function calls. Works with screen‑share and other visual inputs.

Sparrow‑1 (turn‑taking): Transformer‑based conversation timing that adapts to human rhythm (tone, pauses, overlap). Delivers fluid back‑and‑forth and ultra‑fast responsiveness.

Capabilities and platform primitives

Perception (vision): Real‑time reading of facial cues, environment, and on‑screen content to inform responses, capture emotional/behavioral signals, and trigger actions. (Model: Raven‑1.)

Turn‑taking: Natural conversational flow—when to listen, when to interject, when to continue—so sessions feel human, not “botty.” (Model: Sparrow‑1.)

Knowledge Base (RAG): Fast document grounding for accuracy and recency: upload URLs/PDFs/CSVs; use tags; choose Speed / Balanced / Quality retrieval strategy. Designed for low‑latency, real‑time conversations.

Memories: Persistent memory across sessions so conversations “pick up where they left off.”

Objectives & Guardrails: Structured goals (with measurable completion criteria) plus strict behavioral guidelines—keep conversations focused, safe, and on‑brand. Managed on the Persona layer via UI or JSON.

Function calling: Let AI humans take action (tool invocations, workflows) as part of the conversation. Supported across plans and core to “not just talk—also act.”

Bring your own LLM (BYO LLM): Swap in your preferred language model to control tone, cost, or domain expertise while keeping Tavus’ humanlike presence.

White‑labeling: APIs and endpoints designed so the experience is fully your brand—includes white‑labeled Replica consent flows at higher tiers.

Alpha‑channel video: Output with transparency for compositing onto any background/UI.

Transcripts & recordings: Conversation logs and (optional) recordings for analytics, QA, or compliance (per‑minute pricing where applicable).

Concurrency / streams: How many simultaneous live conversations your plan supports (e.g., 1 → 3 → 15 → custom).

Canonical use‑case nouns (how customers talk about outcomes)

Digital twin: A lifelike, interactive stand‑in for a person (expert, coach, creator) used in real‑time conversations. Common customer phrasing; pair with our preferred “AI human” in public copy.

Expert clone: A digital twin of a subject‑matter expert to scale coaching, training, or guidance. (Examples across coaching, sales enablement, tutoring.)

AI interviewer / mock interviews: Face‑to‑face practice with feedback and scorecards; a flagship CVI use case.

AI companion: Humanlike support for wellbeing and daily check‑ins (e.g., senior care, loneliness mitigation) with lifelike presence and low latency.

AI tutor / coach: Role‑play, upskilling, and personalized instruction that adapts in real time to learner signals. This term is used to describe a wide range of applications from B2B internal coaching with tight guardrails, to consumer-facing "study buddy" style RAG use cases.

AI concierge / kiosk: Humanlike service in physical or embedded contexts (hotels, drive‑thru, help desks) with face‑to‑face guidance.

AI eCommerce assistant: Product discovery, fit advice, and guided buying with visual perception and on‑brand presence.

‍