Animation sequence frame generation method and system based on diffusion model

The invention relates to the technical field of game animation picture processing, in particular to an animation sequence frame generation method and system based on a diffusion model, and the method comprises the steps: obtaining a to-be-stylized first picture, carrying out the image preprocessing of the first picture, carrying out the tag backstepping of the picture through an SD plug-in tag, and training a Lora model through the picture; obtaining a sequence frame which needs to generate an action, and carrying out stylization processing on the sequence frame through a Lora model to obtain a second picture; processing the sequence frame through a Contro Net model to obtain a third picture; and combining the second picture and the third picture to generate an available sequence frame. The stylization of picture output is controlled, picture generation is controlled through Lora to ensure that the style of the final result corresponds to the game requirement, and training resources can be greatly saved by us