It looks like a recent ComfyUI update has broken how the primitive string for the prompt is passed into the subgraph in the v4.0 workflow. I'll try to fix this as soon as I can.
This is a simple workflow with refiner for the GonzaLomo and MoP SDXL models.
Description
The only real difference between this version and v3.0 is the addition of the Self-Attention Guidance node on the base image generation. This is especially useful with the MoP (Mixture of Perverts) v2.0 model.
FAQ
Comments (1)
Thank you for sharing this - nice smooth workflow