Every story we've covered involving video-deepfakes.
AI startup Tavus says its new conversational video model, Griffin-Lite, convinced 26 of 54 people (48%) in one-minute live video calls that they were speaking to a real human. The company reports Griffin-Lite tops NVIDIA's VideoFDB generation benchmark and is being tested only by select trusted users while Tavus develops safety safeguards.
No stories here yet.