The paper introduces GaussianFlow, a novel method for generating and synthesizing 4D content using 3D Gaussian Splatting. The key contribution is the introduction of Gaussian flow, which connects the dynamics of 3D Gaussians to pixel velocities in consecutive frames. This allows for direct supervision of 3D Gaussian dynamics using optical flow, improving the quality and realism of 4D content generation and novel view synthesis. The method addresses the under-constrained nature of 4D Gaussian fields by leveraging photometric loss and generative models, while also resolving common issues such as color drifting. Extensive experiments on the Consistent4D and Plenoptic Video datasets demonstrate the effectiveness of GaussianFlow, showing superior results in terms of visual quality and realism compared to existing methods. The method is particularly effective for scenes with rich and dynamic motions, making it a significant advancement in 4D content creation.The paper introduces GaussianFlow, a novel method for generating and synthesizing 4D content using 3D Gaussian Splatting. The key contribution is the introduction of Gaussian flow, which connects the dynamics of 3D Gaussians to pixel velocities in consecutive frames. This allows for direct supervision of 3D Gaussian dynamics using optical flow, improving the quality and realism of 4D content generation and novel view synthesis. The method addresses the under-constrained nature of 4D Gaussian fields by leveraging photometric loss and generative models, while also resolving common issues such as color drifting. Extensive experiments on the Consistent4D and Plenoptic Video datasets demonstrate the effectiveness of GaussianFlow, showing superior results in terms of visual quality and realism compared to existing methods. The method is particularly effective for scenes with rich and dynamic motions, making it a significant advancement in 4D content creation.