metadata

sdk: gradio
sdk_version: 5.0.2
title: Meissonic Flow

Meissonic: Revitalizing Masked Generative Transformers for Efficient High-Resolution Text-to-Image Synthesis

Paper | Model | Code

Introduction

Meissonic is a non-autoregressive mask image modeling text-to-image synthesis model that can generate high-resolution images. It is designed to run on consumer graphics cards.

Prerequisites

git clone https://github.com/huggingface/diffusers.git
cd diffusers
pip install -e .

Usage

python inference.py

Citation

If you find this work helpful, please consider citing:

@article{bai2024meissonic,
  title={Meissonic: Revitalizing Masked Generative Transformers for Efficient High-Resolution Text-to-Image Synthesis},
  author={Bai, Jinbin and Ye, Tian and Chow, Wei and Song, Enxin and Chen, Qing-Guo and Li, Xiangtai and Dong, Zhen and Zhu, Lei and Yan, Shuicheng},
  journal={arXiv preprint arXiv:2410.08261},
  year={2024}
}