A Multi-Stage Approach to Achieving Image Consistency in a Zero Shot Character Generation in D&D domain

In the evolving landscape of computer graphics, the pursuit of consistency in image generation remains a compelling challenge. This paper delineates a comprehensive methodology that harnesses the capabilities of the Stable Diffusion (SD) model, an adept open-source image generation tool, to generate consistent and high-fidelity imagery. Beginning with the training of a generalized art style, our research journeyed through specific character generation, the creation of exhaustive character sheets, and culminated in the generation of consistent character images across various poses.