Tavus
API for real-time AI video agents (PALs) that see, hear and talk back on a live call, plus async avatar video.
How the 7.2 was reached
- Quality of output6/10
Independent benchmark (NVIDIA VideoFDB) and a study with 48% mistaking it for human are strong, but a review calls the voice robotic; no heavy adoption evidence. orcarouter.ai
- Ease of use8/10
Free developer and PALs tiers, no-code PAL Maker, and quickstarts plus deployment guides for web, mobile, Meet, Zoom and Teams. docs.tavus.io
- Pricing value7/10
Free tiers for both account types (15 min calls, 25 min developer video); paid plans from $20/mo to $397/mo plus pay-as-you-go, which is harder to predict. tavus.io
- API and integration quality8/10
Documented CVI and Developer APIs with several models, webhooks, tool calling, connectors, and Meet, Zoom and Teams integrations; SDKs in several languages not shown. docs.tavus.io
- Solves the problem it claims to solve7/10
Independent customer cases (Final Round AI 100K+ users, CareFlick) and the human-likeness study support the claim; reviewer notes of robotic voice leave a small gap. onyxranked.com
7.2/10 is the mean of the 5 criteria that apply. Scored on the five-criteria rubric v1, 2 October 2026. How scoring works
Pricing
- 25 conversational video minutes/month
- 5 minutes of video generation/month
- 1 concurrent conversation
- Stock faces only
- 100 conversational video minutes/month, then $0.37/min
- 10 minutes of video generation/month, then $1/min
- 3 concurrent conversations
- 3 custom faces/month ($65 each after)
- 1,250 conversational video minutes/month, then $0.32/min
- 100 minutes of video generation/month, then $0.90/min
- 10 concurrent conversations
- 7 custom faces/month ($40 each after)
- Volume discounts on minutes
- Custom concurrency
- Unlimited custom faces
- White-labeling and SLAs
Key Features
- Conversational Video Interface: live WebRTC video calls with an AI agent (a PAL)
- Phoenix face rendering, Raven perception (camera, tone, screen share) and Sparrow turn-taking models
- Bring your own LLM, tool calling, MCP connectors, knowledge base, memory stores, objectives and guardrails
- PALs can join Google Meet, Zoom or Microsoft Teams calls from a calendar invite
- Custom faces trained from a short video or a single image
- React component library, embeddable widget, LiveKit and Pipecat integrations, webhooks
Pros & Cons
Pros
- One API covers face, voice, perception, turn-taking and the call itself
- You can swap in your own LLM and tools, so the agent's brain stays yours
- Published per-minute pricing and a free tier to test on
- Docs are deep, with an OpenAPI spec, an MCP server and a CLI
Cons
- Per-minute cost adds up fast for long or high-volume calls
- Concurrency is capped low on self-serve plans (1, 3 or 10 live calls)
- Realism raises disclosure and consent duties you have to design for
- Naming changed from persona/replica to PAL/face, so older tutorials are confusing
Tavus is the most complete API for putting a live AI face on a video call while keeping your own LLM, tools and knowledge behind it. At $0.32 to $0.37 a minute, with 3 or 10 concurrent calls on the self-serve plans, it suits high-value conversations better than high-volume ones.
Try Tavus →Added to scored.tools on
Competitors to Tavus
Other tools in the video category worth comparing.
HeyGen
7.0/10Create professional AI avatar videos in minutes without cameras or studios.
Synthesia
7.6/10Create professional AI avatar videos from text in 140+ languages without cameras, actors, or studios.
ElevenLabs Conversational AI
6.8/10Deploy human-like voice and chat AI agents with low latency in 70+ languages — no IVR menus, no scripts.