The present invention relates to a
system that automatically generates 2D animations based on text and embeddings, maintaining visual and
semantic consistency by combining model IDs and embeddings of characters, backgrounds, and motions. Information regarding scenes, characters, backgrounds, actions, and dialogue is extracted from input text content through
natural language processing (NLP) and semantic analysis. By
combining character embeddings and motion embeddings with model IDs, background images are synthesized to automatically generate
animation scenes while maintaining consistency in the appearance and movement of the characters.
Adaptive learning based on
user feedback is performed to continuously improve the generation performance of the
system, and post-production work is automated to maximize production efficiency. By enabling the efficient and consistent production of 2D animations in large-scale user environments, this
system reduces production costs and time and supports non-experts in easily creating animations.