SD-Turbo

[flynnamy]

It seems that it could finish inference in 4-8 steps, right? Please update about SD-turbo! @IceClear

[IceClear]

Hi. The details of the command has been updated. Feel free to have a try :)

[CCpt5]

FYI, while it was said the SD-Turbo model does -not- work with the 768 resolution SD2.1 model, I can confirm that this new "Trajectory Consistency Distillation" LoRA model does. Can complete inference in 4 steps (they recommend setting cfg to .3).

https://huggingface.co/h1t/TCD-SD21-base-LoRA/blob/main/README.md

Tested w/ the Comfy node.

[IceClear]

FYI, while it was said the SD-Turbo model does -not- work with the 768 resolution SD2.1 model, I can confirm that this new "Trajectory Consistency Distillation" LoRA model does. Can complete inference in 4 steps (they recommend setting cfg to .3).

https://huggingface.co/h1t/TCD-SD21-base-LoRA/blob/main/README.md

Tested w/ the Comfy node.

Thx！ I will have a look when I am available :)