FLUX - Realistic Amateurs ComfyUI Workflow
This workflow is used to generate infinite realistic amateur women using FLUX, Loras, and wildcards!
Share your best creations!
Feel free to adjust the loras used - and wildcards to your preference - but make sure to share your best results and suggestions for improvements here. Enjoy!
Motivation
I've received a few messages recently about my realistic amateur photos and wanted to share my current progress in making a wildcard realistic prompt for making a wide variety of different characters.
The download includes a ComfyUI workflow - and a wildcards.yaml file used to generate randomized aspects of the image!
Models (Install/location)
Acorn Is Spinning FLUX (ComfyUI/models/unet)
VAE (HuggingFace login required) (ComfyUI/models/vae)
Text Encoder (ComfyUI/models/clip)
Clip_l (ComfyUI/models/clip)
Loras (ComfyUI/models/loras):
Loras are entirely optional. The ones below I've found to be good quality to add either style or substance to your images.
Nudity (Choose 1)
[Flux/Pony] Perfect Full Round Breasts & Slim Waist
Style (Choose 1)
Utility
ComfyUI Custom Nodes
Use ComfyUI Manager to install missing custom nodes and restart ComfyUI. List below
Wildcards
For more information on making and adjusting the wildcards, please refer to the wiki page.
Changelog
V1.4
Remove most loras from the default workflow.
Incorporate new FLUX findings by adding a file name and path to the prompt, resulting in a very candid amateur look.
Modify the path name for interesting results!
Updated hair wildcards
V1.3
Workflow overhaul to now use the Image Saver nodes which help with automatic LORA and model detection. Read the notes in the workflow for more information.
Designed for speed - Wildcard Encode only happens for the LORA section which is then combined with the Wildcard prompt.
Started the process of pruning clothing - only keeping what works.
V1.2
Less is more.
Removed detailed face prompts as with testing they didn't seem to have a major impact on the result
Added SameFace Fix [Flux Lora] to avoid 1girl faces
Added FLUX - Female Anatomy for better breast size variation
Removed locations - feel free to add as you see fit.
Simplified the style section of the prompt. Feel free to swap out n30nfilm with whatever style you like!
V1.1
Added more background settings, clothes, and expressions to the wildcards lists.
Shortened the default prompt to be more clear and concise which helps with swapping style LORAs.
Added style lora Vi-FluxFinePortrait
Added nudity lora [Flux/Pony] Perfect Full Round Breasts & Slim Waist
Description
Wildcards:
Added more background settings.
Simplified part wildcards to be more extensible in the future.
Workflow:
Added more loras for variety in style and body composition
FAQ
Comments (18)
in your prompts you have the trigger word for the Midjourney meets FLUX lora, are you also using that lora in your images or just the two you listed in the description?
I've been experimenting with the same prompt/ workflow and different style loras. The loras I have linked in the descriptions are ones that I've tested and liked. I've also been adding some to the Suggested Resources section.
This workflow is definitely meant as a template - so feel free to mix and match style loras that you're interested in!
where are the worflows and the loras? there are other loras in the generations.
i think i figure the gist of it, but where does the JSON file go? i see where the wild card yaml file goes.
@Noob_ee Sorry - I've been on vacation the past couple days. You should be able to launch ComfyUI, then load the .json file as a workflow.
so far, i got most of the workflow working, just hunting down the loras.
@Noob_ee In the newest version - I linked the loras used in the note in the left hand side. All of them should be in the description above too^
@prompt_bit_sorcerer each image has their own set of loras, but i have been on the hunt and found most of the loras
great work, thank you for all your efforts.
fantastic workflow , firstly congratulations and thanks for sharing, i am still a noob when it comes to creating workflows , i tried and failed to add an upscaler to youra , do you have plans to add an upscaler to your next iteration of this workflow?
A good possible addition in the future! I generally don't use one because of the way I wildcard prompt. I don't want to upscale a bunch of images I don't plan on keeping (not every one is a winner when going random). I suppose I could make it optional. That said, I did start this workflow from the base of this workflow - which I'm sure you could easily copy pieces from - or port this to - https://civitai.com/models/642589/flux-modular-workflow-original-or-gguf-flux-model-latent-noise-injection-face-detailer-faceswap-and-ultimate-sd-upscaler
Producing beautiful images with this workflow, most realistic results I've had with Flux...only I'm not experienced with wildcards, is there a resource that can explain how I change the prompt you have include in this workflow and still draw on your wildcard yaml ???
in the wildcard prompt and guidance box , leave everything within the ( ) and { } you can change the normal text to what you want ( new to wildcards myself but just tried and it works )
@GG_System Cheers I'll have a play..
Exactly. Think about just changing a part of a sentence. The sky was {blue|red|yellow}. The brackets and pipes are just saying "Hey, here's the part to make random". Here's the wiki, where you can learn more about how to say choose 1-2 random things: https://github.com/ltdrdata/ComfyUI-extension-tutorials/blob/Main/ComfyUI-Impact-Pack/tutorial/ImpactWildcard.md
@prompt_bit_sorcerer I've added extra .txt folders into the custom wildcard folder and they don't seem to be recognised ??? Your Yaml file is and obviously can be accessed through the drop down in the node...as are the sample prompts but nothing else I'm putting in there ?? Any ideas ??
@goldman I haven't used txt files as wildcards - so not sure on that one. I know in ComfyUI you need to "Refresh" before wildcard changes appear in the UI - maybe a similar problem for you? Personally I would stick with YAML - even if you make multiple files. If you still have trouble you may check the git issues for the plugin to see if someone has a similar problem.
@prompt_bit_sorcerer cheers will do that..