# Generating Pixel Art Character Sprites using GANs

arXiv: https://arxiv.org/abs/2208.06413 · HTML: https://mlsanity.com/paper/2208.06413/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- Builds on 5 of its 17 references that are in the map.
- First posted on arXiv 15 Aug 2022.

## Builds on (most cited first, 5 of 5)

- [Image-to-Image Translation with Conditional Adversarial Networks](https://mlsanity.com/paper/1611.07004/) (2016)
- [Conditional Generative Adversarial Nets](https://mlsanity.com/paper/1411.1784/) (2014)
- [Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks](https://mlsanity.com/paper/1511.06434/) (2015)
- [StarGAN: Unified Generative Adversarial Networks for Multi-Domain Image-to-Image Translation](https://mlsanity.com/paper/1711.09020/) (2017)
- [Generating Gameplay-Relevant Art Assets with Transfer Learning](https://mlsanity.com/paper/2010.01681/) (2020)

## Built on by (most cited first, 3 of 3)

- [SD-πXL: Generating Low-Resolution Quantized Imagery via Score Distillation](https://mlsanity.com/paper/2410.06236/) (2024)
- [Voxify3D: Pixel Art Meets Volumetric Rendering](https://mlsanity.com/paper/2512.07834/) (2025)
- [A Missing Data Imputation GAN for Character Sprite Generation](https://mlsanity.com/paper/2409.10721/) (2024)

## Abstract (arXiv)

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.