Any chance of getting some sort of faceswap node going? Mostly interested in either use the first output to generate multiple other outputs as the same character. I keep trying to find a way to incorperate it after flux generation, but you're updating your workflows so fast I can't keep up. Might as well ask/suggest as an idea.

But this workflow, mate, is really something. It works exactly how I want it with minimal effort and it just keeps getting better and better. Love it!

how do i upload my own lora into the workflow?

服务器不够的小伙伴可以直接用悦灵犀app

# ComfyUI Error Report
## Error Details
- **Node Type:** TaraAdvancedComposition
- **Exception Type:** openai.APIConnectionError
- **Exception Message:** Connection error.
## Stack Trace

What is the solution to this problem?

It's all red, maybe because of ComfyUI's new 0.22 update?

Thanks for this amazing Workflow. I'm using normal prompt with latent noise injection, adetailer and ultimate sd upscaler After some finetuning while using loras it produces great images!

I have one problem: The first facedetailer works well. The mask covers the face. If I understand correctly, the second facedetailer should then improve the eyes, but despite the correct bbox model, it does not recognize the eyes, i.e. no mask of the eyes is recognizable. Therefore, it does not process this region.

What could be the reason for this? Everything is set to standard and these are simple images of people and portaits.

Workflow Ver. 4.

When using some loras (flux_realism, boreal, ...) I get moire patterns with Latent Noice Injection, but only in conjunction with Latent Upscaling. How can I avoid that?

So far this workflow is amazing. Just a question, is it all layered in the sense that when u enable faceswapper, will it take on that result of that picture and work on that? Or do you need change it to img2img and load that image?

I'm having problems installing the following:

• ReActorFaceSwap
• ReActorFaceBoost

ComfyUI manager doesn't seem to support them. Any fix please? I tired manual install it didn't work.

Would love if you add another option for true img2img in addition to the florence img2prompt2img. Unless I'm missing something and it's somewhere in the WF?

I believe some update has "broken" the denoise function.

• Exception Message: BasicScheduler.get_sigmas() missing 1 required positional argument: 'denoise'

Getting this error in the sampler 1 node in the latent injection group,
Here img2img and latent injection are enabled.

"RuntimeError: cannot reshape tensor of 0 elements into shape [1, 0, 3, 24, -1] because the unspecified dimension size -1 can be any value and is ambiguous"

I just came back from a 1-week vacation, and I see many messages about this error in my workflow.  

I guess there has been an update (in the nodes or in ComfyUI) that broke the workflow.

I will do a nice backup first of all my ComfyUI folder without the updates, then I will update everything and start to check the errors and try to fix them... Stay tune and sorry for the inconveniece.

Gorgeous  workflow, i have only a question, how can i set the ending group the number 8 to use all previous one and keep the replaced face? because if i active all groups, at the end i have the one that comes from group 7 not even 10, thank you

i am confused, if i activate the face swap nodes then the detailer is still working with the non swapped face, how i can have the face swap to work at the end so that the detailer consider the swapped face and not a random one?

And how you are supposed to make the changes with the inpaint, in none of the nodes there is a prompt for it

I m starting from zero , so I'm missing quite a few nodes, how can I install them? the civitai link is not working anymore