4D-fy is a text-to-4D generation method that employs mixed-fraction distillation sampling, combining supervised signals from various pre-trained diffusion models to achieve high-fidelity text-to-4D scene generation. Its approach parameterizes 4D radiance fields with neural representations, utilizing static and dynamic multi-scale hash table features. It then renders images and videos from the representations using volume rendering. Through mixed-fraction distillation sampling, 4D-fy first optimizes the representation using gradients from a 3D-perceptual text-to-image model (3D-T2I), then refines the appearance by incorporating gradients from a text-to-image model (T2I), and finally enhances the scene's motion by incorporating gradients from a text-to-video model (T2V). 4D-fy can generate 4D scenes with captivating appearance, 3D structure, and movement.