Workflow for V-pred models and loras


Updated:

I posted my comfyUI workflow which you can use on tensorart: https://tensor.art/workflows/923317298096284260

It supports rescale CFG and has forge-like workflow (txt2img -> 2-pass highres fix -> facedetailer (face and eyes).

You can bypass rescale CFG node if you will use it with epsilon models instead of V-pred.

It can be little complicated but I can help. I am not so deep in comfyUI because I am forge / A1111 enjoyer but I did this workflow by myself from zero.

Also I added facedetailer workflow for people who want to improve faces on their pics (For example if you did it on civitai which has not so good facefix):

https://tensor.art/workflows/923327030492139408

0