The Brazilian game industry is mostly composed of small studios operating with limited financial resources and small teams. In this scenario, tasks that demand repetitive manual effort become potential bottlenecks in the development flow. Among the pixel art processes in 2D games, the creation of charsets stands out, which are structures formed by multiple animation frames of the same character. This is a particularly costly process, as any alteration to a sprite requires rework on several associated frames to maintain visual consistency. Faced with this challenge, this work proposes the development of a generative adversarial network (GAN) to assist in the automatic generation of charsets, reducing work time and consequently lowering production costs. The objective is to develop a GAN model capable of generating a complete charset in the RPG Maker engine standard, from a single input sprite. The proposed solution consists of implementing an adapted version of the Pix2Pix architecture, configured to generate the character set from a single sprite. After training, the model is evaluated using FID and L1 Loss metrics, complemented by a visual analysis of the results. It is expected to demonstrate that the "one-to-many" approach is viable and can act as a support tool for independent developers, aiming to promote productivity gains.