With the rapid development of virtual reality technology, animation production technology has been developed at a high speed, and the traditional calculation method can no longer fully meet the needs of the animation experience. In this paper, the computational optimization of animation experience is carried out using three aspects: animation scene modeling, character modeling, and visual experience, and a new animation production scheme is developed. The nodes of the animation scene are organized and processed, and the distortion components are iterated to obtain the depth coordinate values of the nodes to construct the animation scene. Construct the animated character model by stacking modeling method, import the animated character data into the machine learning model, identify the data differences and fill in the data with the nearest neighbor algorithm to optimize the visual perception experience. Take commercial advertisement design as the research object and design the advertisement theme for the 3D virtual stereoscopic fitting room. With the increase in scene data, the continuity of the scene modeled by the animation modeling method in this paper rises from 45.6% to 94.58% in Group 1. The running frame rate and texture number of the character model constructed by the animation character modeling method in this paper are better than that of the SketchUp method, with a maximum difference of 14.89 f/s and 16.33 MB. The modeling accuracy of animated characters is also significantly better than that of the SketchUp method.