As of: 2026-11-19 22:11 CET. Researched and written by AI agents (Claude Opus 5.5 in Claude Code), with a human editor. How we work: https://postcutoff.com/about/ Canonical page: https://postcutoff.com/m/tavus-griffin/ # Tavus Griffin (Human Interaction Model) Tavus, Griffin family. Status: preview. Type: video generation. Released: 2026-10-01. ## How to call it | Provider | Model id | Endpoint | Docs | |---|---|---|---| | Tavus (research preview) | | | https://www.tavus.io/griffin | ## Pricing - Price not published. - Last checked: 2026-10-02 ## Specs - Input: video, audio - Output: video, audio - Context window: not published - Max output: not published - Open weights: no - Licence: proprietary ## Where it stands - Newer video generation models in the registry since its release: 3: - Gemini Omni 1.2 Flash (2026-11-05) https://postcutoff.com/m/gemini-omni-1-2-flash/ - FLUX 3 Max (2026-10-19) https://postcutoff.com/m/flux-3-max/ - Kling 4.0 (and 4.0 Flash) (2026-10-15) https://postcutoff.com/m/kling-4-0/ - Closed weights. ## What stands out - Full-duplex face-to-face video conversation (first): Video-to-video model that sees and hears the user while speaking, interrupts and takes interruptions, and reacts to gestures and objects in real time; Tavus calls it the first 'Human Interaction Model'. Source: https://www.tavus.io/griffin - 'Video Turing test' result (vendor study): 48% of 54 participants in 1-minute blind live calls judged it human (Tavus's earlier stack ≤2%); NVIDIA VideoFDB generation 3.83/5 vs human 3.92; median latency ~1.9 s. Source: https://www.tavus.io/griffin ## Notes Research preview; no public API id or pricing found. Benchmarks: Tavus's own study plus NVIDIA's VideoFDB leaderboard (Griffin-Lite). Model page: https://postcutoff.com/m/tavus-griffin/