Generating Pixel Art Character Sprites using GANs

TL;DR

This paper explores using conditional GANs, specifically Pix2Pix, to automate and assist in creating pixel art character sprite sheets for game development, aiming to reduce manual effort.

Contribution

It introduces a Pix2Pix-based architecture tailored for generating character sprites in different poses, demonstrating its potential in pixel art sprite sheet creation.

Findings

Models can generate sprites close to ground truth.

Varying degrees of generalization observed across datasets.

Quantitative analysis with FID supports visual results.

Abstract

Iterating on creating pixel art character sprite sheets is essential to the game development process. However, it can take a lot of effort until the final versions containing different poses and animation clips are achieved. This paper investigates using conditional generative adversarial networks to aid the designers in creating such sprite sheets. We propose an architecture based on Pix2Pix to generate images of characters facing a target side (e.g., right) given sprites of them in a source pose (e.g., front). Experiments with small pixel art datasets yielded promising results, resulting in models with varying degrees of generalization, sometimes capable of generating images very close to the ground truth. We analyze the results through visual inspection and quantitatively with FID.