aMUSEd: An Open MUSE Reproduction

TL;DR

aMUSEd introduces a lightweight, open-source masked image model for fast, interpretable text-to-image generation, requiring fewer parameters and inference steps than latent diffusion, with the ability to learn new styles from minimal data.

Contribution

It presents aMUSEd, a compact MIM model that enhances text-to-image generation speed and interpretability, and provides reproducible code and checkpoints for large-scale use.

Findings

aMUSEd achieves fast image generation with 10% of MUSE's parameters.

It requires fewer inference steps than latent diffusion models.

The model can learn new styles from a single image.

Abstract

We present aMUSEd, an open-source, lightweight masked image model (MIM) for text-to-image generation based on MUSE. With 10 percent of MUSE's parameters, aMUSEd is focused on fast image generation. We believe MIM is under-explored compared to latent diffusion, the prevailing approach for text-to-image generation. Compared to latent diffusion, MIM requires fewer inference steps and is more interpretable. Additionally, MIM can be fine-tuned to learn additional styles with only a single image. We hope to encourage further exploration of MIM by demonstrating its effectiveness on large-scale text-to-image generation and releasing reproducible training code. We also release checkpoints for two models which directly produce images at 256x256 and 512x512 resolutions.