Griffin
A Human Interaction Model released by Tavus, described as the first of its kind. It supports multimodal real-time perception—including facial expressions, gestures, and scene context—and features full-duplex video-to-video interaction, enabling simultaneous observation of and immediate response to conversation partners.
Timeline (last 90 days)
Tavus Griffin provides real-time face-to-face AI video interaction, where the digital human can respond to user conversations instantly.
Tavus Griffin is positioned as an SDK that can be embedded into enterprise websites or applications, targeting B2B teams.
Griffin can simultaneously see, hear, and speak—observing and reacting at the same time rather than interacting in turns.
Tavus has released Griffin, positioning it as the world's first Human Interaction Model.
Griffin can respond to pauses, facial expressions, body gestures, and scenes in front of the camera during a conversation.
Griffin ranked 4th in the Artificial Intelligence category on the Product Hunt daily leaderboard with 99 upvotes.
In a real-time video call lasting one minute, 48% of participants believed they were talking to a real person.
Griffin received only 2 comments on Product Hunt, providing a limited sample of community feedback.
Tavus unveiled Griffin on October 1, calling it the first human interaction model, integrating real-time perception, conversation, speech, and video generation into a full-duplex interaction system; Griffin Lite is currently available as a research preview to select early testers only.
In a one-minute video call test, 26 out of 54 participants (approximately 48%) mistook Griffin's counterpart for a real human; Griffin outputs 720P video in 320-millisecond chunks, with an average audio-to-video latency of approximately 0.43 seconds on H100.
3 more timeline events
All Facts (13)
Tavus Griffin provides real-time face-to-face AI video interaction, where the digital human can respond to user conversations instantly.
50%UnverifiedTavus Griffin is positioned as an SDK that can be embedded into enterprise websites or applications, targeting B2B teams.
50%UnverifiedGriffin can simultaneously see, hear, and speak—observing and reacting at the same time rather than interacting in turns.
50%UnverifiedGriffin can respond to pauses, facial expressions, body gestures, and scenes in front of the camera during a conversation.
50%UnverifiedGriffin ranked 4th in the Artificial Intelligence category on the Product Hunt daily leaderboard with 99 upvotes.
50%UnverifiedIn a real-time video call lasting one minute, 48% of participants believed they were talking to a real person.
50%UnverifiedGriffin received only 2 comments on Product Hunt, providing a limited sample of community feedback.
50%UnverifiedIn a one-minute video call test, 26 out of 54 participants (approximately 48%) mistook Griffin's counterpart for a real human; Griffin outputs 720P video in 320-millisecond chunks, with an average audio-to-video latency of approximately 0.43 seconds on H100.
50%UnverifiedDeepMind的Griffin将基于线性递归单元LRU的循环层与局部注意力层交替堆叠
50%UnverifiedHydraNet-VSM的并行融合架构理念与NVIDIA的Hymba和DeepMind的Griffin架构相似
50%Unverified2024年涌现了多个线性或近线性复杂度序列模型，包括DeepMind的Griffin、AI21 Labs的Jamba、HGRN2和Stanford的Based
50%UnverifiedTavus has released Griffin, positioning it as the world's first Human Interaction Model.
50%UnverifiedTavus unveiled Griffin on October 1, calling it the first human interaction model, integrating real-time perception, conversation, speech, and video generation into a full-duplex interaction system; Griffin Lite is currently available as a research preview to select early testers only.
50%