Tavus Freemium

What it is

Tavus is a San Francisco company that builds "AI humans": real-time conversational video avatars that see, listen and respond. Its Conversational Video Interface (CVI) bundles the LLM, speech, perception and rendering into one API and no-code platform. It also makes asynchronous personalised videos from a custom replica.

Key features

• Phoenix-4.5 full-face rendering with emotion control
• Raven-1 perception and vision, and Sparrow-2 turn-taking
• Knowledge base (RAG), memories, objectives, guardrails and function calling
• Google Meet and Zoom integration, Pipecat and LiveKit connectors
• Stock faces (10-100+) and custom Face slots trained on a person
• Async video generation minutes for personalised outreach videos
• 42+ languages, 1080p, 24 kHz audio, bring your own LLM

Models inside

Tavus's own Phoenix-4.5, Raven-1 and Sparrow-2 models, plus a fine-tuned LLM or your own. No third-party generative models from the list are documented.

Pricing and rights

As of September 2026: Free includes 20 conversational minutes a month, stock faces and 5-minute sessions, with a watermark. Starter is $22/month for 60 minutes, 1 custom face, no watermark and 5 async video minutes. Builder is $59 for 175 minutes and 3 custom faces, with overage at $0.35 a minute. Growth is $397 and Business $975 for 1,300 and 4,000 minutes. Enterprise adds SOC 2/HIPAA and zero data retention. Tavus's terms say you own generated media files. You must get explicit consent and written releases from anyone whose likeness, voice or biometrics you use, and you may not impersonate anyone without consent.

Who it is for

Product teams and developers building AI tutors, sales or support agents, interviewers and personalised video outreach.

Verdict

Tavus leads on real-time, two-way avatar conversation, with a well-rounded developer stack. It's more of an infrastructure product than a video editor, sessions are billed by the minute, and creators who want finished marketing videos will find it indirect.