Unwrapped

Teardown · tavus

TAVUS

TAVUS

CategoryVideo AvatarsLast round · $18M · 2024Site ↗
  • Sequoia Capital

Custom avatar models + LLM APIs + real-time rendering pipeline.

01

Public data / API layer

CR
Customer-provided replica training videoYours
CK
Customer knowledge base (CSVs, PDFs, websites)Yours
TS
Tavus stock replica library (100+ pre-trained avatars)Licensed
Frontier LLM corpora (via API)
Frontier LLM corpora (via API)API

Internal replication score

Medium
0.36

Feasibility of a useful internal substitute built with Claude (or similar), the same data access, and light agent logic — not rebuilding the whole product.

IRS = 0.30·D + 0.25·L + 0.20·O + 0.15·R + 0.10·Sthis record · 36%
  • D

    Data accessibility

    weight 0.300.30
    • 1.0mostly customer-owned / public / standard third-party sources
    • 0.5mixed accessibility
    • 0.0hard-to-access or proprietary source layer
  • L

    LLM substitutability

    weight 0.250.20
    • 1.0mostly retrieve / prompt / cite / summarize / classify / compare
    • 0.5mixed standard + custom behavior
    • 0.0strongly custom model behavior (fine-tunes on proprietary data, etc.)
  • O

    Output simplicity

    weight 0.200.50
    • 1.0straightforward internal work product (memo, list, reply, SQL query)
    • 0.5moderately specialized
    • 0.0highly specialized (e.g. FDA-graded clinical text)
  • R

    Review / risk tolerance

    weight 0.150.60
    • 1.0internal use with human review is acceptable
    • 0.5moderate risk
    • 0.0very low tolerance for error (e.g. external legal filings)
  • S

    Surface complexity

    weight 0.10inverse — higher means less surface dependence0.30
    • 1.0a simple internal shell is enough
    • 0.5polished workflow matters somewhat
    • 0.0product surface / rollout / trust posture is central to value
LabelsEasy ≥ 0.67Medium ≥ 0.34Hard < 0.34

Missing factor rows use heuristics from wrapper scores. Editorial heuristic, not investment advice.

Build it yourself→

Recreate the workflow inside your org.

Internal build

Build it yourself

Same avatar synthesis + LLM API + WebRTC stack — far harder rendering pipeline, emotion models require deep ML.

Internal use only. Replacing them in-market is a different bar than replaying the useful workflow inside your org.

01 · Connectors & flow

CR
Customer-provided replica training video
CK
Customer knowledge base (CSVs, PDFs, websites)
TS
Tavus stock replica library (100+ pre-trained avatars)
Frontier LLM corpora (via API)
Frontier LLM corpora (via API)

Internal build map

Data in

Connectors
Connectors

Agent layer

Planner
Tools + retrieval
Reasoning model

Logic

LLM API
custom rendering model
emotion detection
turn-taking
WebRTC streaming
custom weights (avatar + behavior)

Outputs

Internal search
Answer
Citations

02 · Claude / agent prompt

Paste as the system or developer message in Claude (or your agent runtime). Scroll to read; Copy grabs the full text.

Claude / agent prompt

// Real-time conversational video agent You are a conversational assistant appearing as a video avatar inside [YOUR_COMPANY]. You help [YOUR_TEAM] using ONLY materials the user is allowed to access: uploaded knowledge base files (CSVs, PDFs, websites), conversation history, and function call results. ## What you must do 1. Retrieve first: Check knowledge base and conversation memory before answering. Cite sources when available. 2. Respond naturally: Use conversational timing. Do not interrupt. Acknowledge what the user says before pivoting. 3. Surface conflicts: If knowledge base entries contradict, state both and ask the user for clarification. 4. Scope: Stay within the uploaded knowledge base and conversation history. Do not fabricate information. ## What you are not Not a replacement for human judgment. Internal use only. Conversations should be reviewed before external use. ## Refusal If asked for information outside the knowledge base, say you do not have access to that material and ask the user to upload it or provide more context. ## Safety Internal conversational assistant. Human review required before external deployment. Do not store sensitive data in memory without user consent.

03 · Result

Walk me through the sales onboarding process for enterprise accounts.
Sales_Playbook.pdf

Based on the uploaded Sales_Playbook.pdf, enterprise onboarding has four stages: contract signature, kickoff call, admin setup, and first-use training.