The models are not mine originally, I upload them here for convinience and to make them easy to find.

Please post your results in the gallery, and feel free to comment if you found out that some things works better or worse than others 😁

Models:

PixArt-Sigma-XL-2-1024-MS:

Works best to generate in 1 megapixel (i.e 1024x1024 , 1280x768, 896x1152 etc)

Steps: ~15
cfg: ~2-7
Sampler: dpm_2
Scheduler: normal, karras

VAE: SDXL

PixArt-Sigma-XL-2-2K-MS:

Works best to generate in 1 megapixel (i.e 2048x2048, 2560x1536, 1792x2304 etc)

Steps: ~15
cfg: ~2-7
Sampler: dpm_2, dpmpp_2m_sde
Scheduler: normal, karras

VAE: SDXL

PixArt-LCM-XL-2-1024-MS

Steps: ~5
cfg: ~1.1-1.5
Sampler: LCM
Scheduler: normal
VAE: SDXL, SD 1.5
Use with: ModelSamplingDiscrete



@misc{chen2024pixartsigma,

title={PixArt-\Sigma: Weak-to-Strong Training of Diffusion Transformer for 4K Text-to-Image Generation},

author={Junsong Chen and Chongjian Ge and Enze Xie and Yue Wu and Lewei Yao and Xiaozhe Ren and Zhongdao Wang and Ping Luo and Huchuan Lu and Zhenguo Li},

year={2024},

eprint={2403.04692},

archivePrefix={arXiv},

primaryClass={cs.CV}