A workfllow for Stable Cascade Text2Image Basic function.

You will see three LORA loaders here, but I muted them because there are no LORA made base on Stage_C U-net model now [02/21/2024].

Adjust [Compression] level in SC-EmptyLatents from 28 to more higher can save time, but inhibit the quality of image. However, compression level lower than 28 does not means improve the quality, but always increasing the burden on the GPU.

Before you Start, here are FOUR models you need to download from Stable-Cascade official huggingface website.

They are [Stage_a] [Stage_b] [Stage_c] and [Text_Encoder]

Put [Stage_a] into ComfyUI/models/vae.

Put [Text_Encoder] into ComfyUI/models/clip.

Put [Stage_b] and [Stage_c] into ComfyUI/models/unet

And make sure your ComfyUI version is the newest

If you done all above, load my workfllow then enjoy the generation.

[Estimation from PAseer]

Pros: Unknown...maybe quicker ?

Cons: many .... low quality and not flexible