What is Tavus?
Tavus is API-first infrastructure for shipping AI video agents that hold real-time, face-to-face conversations rather than pre-rendered avatar clips. Its Conversational Video Interface (CVI) pipeline handles visual perception, turn-taking, speech recognition, the LLM, text-to-speech, and photorealistic real-time face rendering (Phoenix) end to end, so developers plug in their own LLM/voice/knowledge base and get a live video agent without building the pipeline themselves.
Key features
- End-to-end real-time pipeline: perception, turn-taking, STT, LLM, TTS, and face rendering
- 100+ stock avatars, or train a custom face from 2 minutes of video
- Sub-500ms average speech-to-video latency for natural conversation
- Swap in your own LLM, TTS provider (ElevenLabs, Cartesia, etc.), or knowledge base
- Can join live Google Meet calls as a participant, on camera
- SOC2, HIPAA, and GDPR compliant, with white-label options for enterprise
How to get started
- Sign up for free API credits, no credit card required
- Create a PAL (behavior, knowledge, pipeline config) via the API
- Create or select a Face - a stock avatar or one trained from your own video
- Start a Conversation session to link the PAL and Face in a live video call
- Embed the resulting interface in your product using Tavus's UI components
Tavus pricing
Tavus is priced by CVI minutes rather than seats. The free developer tier includes 25 minutes to build and test.
Starter and Growth scale minute allotments and features for production use, with a custom Enterprise tier for white-labeling and volume.
| Plan | Price | Best for |
|---|---|---|
| Free | $0 | Developers building and testing (25 CVI minutes) |
| Starter | $59/mo | Small production use (100 minutes) |
| Growth | $397/mo | Scaling teams (1,250 minutes) |
| Enterprise | Custom | White-label, volume discounts, dedicated support |
Our take: This is infrastructure for developers, not an end-user video tool - the closest comparison is D-ID or HeyGen's API tier, not their consumer apps.
Prices verified 2026-09. Plans change often — confirm on the official site above before you buy.
Use cases
- Customer-support or sales agents that appear on video and respond in real time
- Interactive training or onboarding experiences with a live AI presenter
- Products that need an AI representative to join a live video meeting
Tavus vs. alternatives
| Tool | How it compares |
|---|---|
| Tavus | Developer API for building real-time, face-to-face AI video agents that see, listen, and respond live. |
| D-ID | D-ID offers similar real-time avatar APIs with its own separate pricing and latency profile. |
| HeyGen | HeyGen is more consumer-facing for pre-rendered avatar videos rather than live conversational agents. |
FAQ
Is Tavus free to use?
Tavus's plans: Free tier (25 CVI minutes), Starter $59/mo, Growth $397/mo. Pricing and free-tier limits change over time, so check the official site above for the latest details.
What is Tavus used for?
Developer API for building real-time, face-to-face AI video agents that see, listen, and respond live.
Who is Tavus best for?
Tavus is a good fit for customer-support or sales agents that appear on video and respond in real time, or interactive training or onboarding experiences with a live AI presenter.
What are the alternatives to Tavus?
Commonly compared alternatives include D-ID and HeyGen, see the comparison above for how they differ.
How much does Tavus cost?
Tavus is billed by CVI minutes. Free includes 25 minutes for development. Starter is $59/month for 100 minutes and Growth is $397/month for 1,250 minutes; Enterprise pricing is negotiated directly for white-label and high-volume needs.
Related tools
D
HLast updated: 2026-09 · Reviewed by AIKetra editors · How we evaluate tools · Embed a Featured badge