Workflow I use to create the preview images on my models here on Civitai. It creates the metadata that's compatible with the site's resource tagging so you get accurate info on your images hosted here.
How to use:
Set your seed to random in the first group "Text 2 Image" While all other groups and the last metadata save node is set to bypass, queue and review each image until you get one you like
Then lock in your seed, set the next group in order, face detailer, to "Set Group Nodes to Always". You can bypass any group you like. The seed here is fixed by default, choose another set seed if the face you get is undesirable. Additionally, there are new prompt boxes if you need specific control over the face generated.
When you have the face detailed, you can upscale if you like. Here you can choose a different model to upscale if you need, otherwise choose the same model.
After upscale, you may need to add details to faces for the final image. Works just like the first face detailer.
Finally, you can just right-click save any image in the process and it will contain the workflow and metadata normally found in a comfyui image. Additionally, you can set the final node located in the top-right to not be bypassed to save an image with metadata that Civitai is more compatible with. Here, the original prompt, seed, height and width is automatically piped in for convenience, you should set the model, cfg, sampler and scheduler manually in this node. You can create nodes and pipelines to automate these as well, but this is my workflow.