Traditional animation creation faces problems such as high time consumption and cost, limited creativity due to manual skills, and difficulty in meeting large-scale production needs. This article aimed to optimize the production process, reduce costs, and enhance innovation by introducing artificial intelligence technology, in order to promote the development of the animation industry to a higher level. This article conducted experiments on the decision tree fusion algorithm, and the experimental results showed that the decision tree fusion algorithm performed better than individual logistic regression algorithms and enhanced decision tree algorithms. This study used the MGif dataset to preprocess the data for quality assurance. By introducing multiple elements to initially design animation types, feature extraction and reconstruction of animation types were carried out to obtain richer animation creation content. The fusion algorithm LSTM-GAN (Long Short-term Memory-Generative Adversarial Network) scheme was adopted. The experimental results showed that the design efficiency of the LSTM-GAN fusion algorithm model was 13.75 h, and the animation creation quality reached 87.084. The introduction of long short-term memory generative adversarial networks has shown excellent performance in animation creation, not only effectively improving animation quality, but also shortening the total creation time.