GenFlare is an artificial intelligence model developed by Baidu for video generation, with a primary focus on image-to-video (img2vid) transformation. Released in late 2025, the model is designed to animate static images into short, cinematic video clips featuring realistic motion and high temporal consistency.
The model utilizes temporal diffusion modeling to ensure fluid movement between frames. In its 2.0 iteration, GenFlare reached top positions on industry leaderboards, such as the Artificial Analysis Video Generation Arena, where it was noted for its motion realism and visual quality. It is specifically optimized to handle diverse cultural and architectural contexts, providing high-fidelity results for global and Chinese markets.
Key capabilities include generating smooth 5-to-10 second sequences with low latency. While the model excels at visual motion dynamics, early versions focused primarily on video synthesis without native sound generation.