Z Image De Turbo
by ostris
Open source · 4k downloads · 162 likes
2.8
(162 reviews)ImageAPI & LocalAbout
Z Image De Turbo is a refined version of the Z-Image-Turbo model, designed to enhance its performance while maintaining compatibility. It enables training of LoRAs or further fine-tuning beyond the original model’s limitations, offering greater flexibility for customized applications. Ideal for generating high-quality images with optimized settings (low CFG and reduced steps), it stands out for its ability to deliver precise results while remaining accessible. This model is particularly suited for creators looking to refine or adapt Z-Image-Turbo without relying on the base version, which has not yet been released. Its simplified approach makes it a practical tool for creative or professional projects.
Documentation
Z-Image-De-Turbo
This is a de-distilled version of Tongyi-MAI/Z-Image-Turbo. It was fine tuned on images generated by Tongyi-MAI/Z-Image-Turbo in order to break down the turbo distillation. It can be used to train LoRAs on top of that should remain compatability with the base model, or it can be continued to finetune well beyond what you could do with the turbo model with the ostris/zimage_turbo_training_adapter.
There is a ComfyUI version as well as a diffusers based version in the repo.
Usage
This can be trained on directly without an adapter. It can also be used for inference directly. For inference, use a low CFG (2.0 - 3.0) and 20-30 steps. It seems to work well with CFG normalization as well.
Why Not Just Use the Base Model?
It wasnt released yet, as of the time of this writing, and I am impatient. :)
Capabilities & Tags
diffuserssafetensorstext-to-image
Links & Resources