Z-Image Turbo++ produces high-fidelity images in 2 steps

Arxiv CS.CV (computer vision)

The model is distilled from the 8-step Z-Image Turbo teacher using teacher-aligned end-to-end distillation. It achieves high-quality 2-step generation, addressing the challenge of pushing diffusion distillation to extremely low step counts.

Text-to-imageDiffusion

See 1 more source

[Papers] - Z Image Turbo ++ - High-Fidelity 2-Step Image Generation via Teacher-Aligned End-to-End Distillation3 hours agoCrazy-Repeat-2006Discuss

Arxiv CS.CV (computer vision)·Dongyang LiuRuoyi DuDavid LiuDengyang JiangLiangchen LiQilong WuZhen LiSteven C. H. HoiHongsheng LiPeng Gao

2 days ago