What it does?
This takes any input image and will replace the person/etc. of your choice with a lora, completely automatically! It will also run a face detailer pass to improve fidelity.
Masking and prompting is automatically taken care of!
It is dependent on an LLM. I recommend using a service as Flux already eats quite a lot of VRAM locally. Groq (as of this writing) offers free LLMs in the cloud, and Plush Nodes documentation (link) will help you get connected.
How to use:
Reference notes within the workflow. The 3 simple fields that you typically need to modify are conveniently grouped within the workflow!
Please enjoy this workflow, and feel free to offer suggestions on future enhancements for a potential new version.
Description
See notes within workflow.
FAQ
Comments (5)
There is probably something needed to be downloaded manually, cause for me it gives white pictures. No comfyui errors, but the "Advanced prompt enhancer" is surrounded red. And the "get text" node next to it isn't relative to my loaded picture.
yeah. Concat String (Klinter) needs to be downloaded manually. Not sure if you would bother it though, results I got with this workflow were abysmal.
Advanced prompt enhancer is part of plush nodes. Is Comfy manager not getting missing nodes for you?
@hadeng Ouch. I tested this thoroughly and got great results 85-90% of the time. There were just a few instances where the Lora didn’t conform to the bounding box or left odd shadows/background artifacts but that’s the nature of inpainting typically.
@EnragedAntelope I got all the nodes. Something else is missing.