File size: 1,370 Bytes
b1ac57e 5c5b395 b1ac57e 2420a18 b1ac57e 582580b b1ac57e fdc7e30 d1afff5 b1ac57e 08ff13d b1ac57e cbbe4f7 b1ac57e 2420a18 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 |
---
pipeline_tag: text-to-image
license: apache-2.0
tags:
- Non-Autoregressive
---
# Meissonic: Revitalizing Masked Generative Transformers for Efficient High-Resolution Text-to-Image Synthesis
[Paper](https://arxiv.org/abs/2410.08261) | [Model](https://huggingface.co/MeissonFlow/Meissonic) | [Code](https://github.com/viiika/Meissonic) | [Demo](https://huggingface.co/spaces/MeissonFlow/meissonic)
## 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.
**Note: This is a project under development. If you encounter any specific performance issues or find significant discrepancies with the results reported in the paper, please submit an issue on the GitHub repository! Thank you for your support!**
## Usage
Under Construction. Please check back later.
## Citation
If you find this work helpful, please consider citing:
```bibtex
@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}
}
``` |