>>> If you think my content is worth it, consider donating to my Patreon or Ko-Fi to help fund the training of new LoRA's or porting existing LoRA's over to other base models! <<<
Trigger Phrase (must be included in the prompt for the full effect)
v2: 'with skin with textured detail'
v1: real skin
Recommended inference settings/workflow
use the default "9B Text-to-Image" ComfyUI workflow (found here) with an added "Load LoRa" node and the following settings (which are the default settings for the model as per its HuggingFace page):
euler (sampler), 50 steps, cfg 4, 1 megapixel resolution (i.e. 1024x1024 for a 1:1 aspect ratio)
How to prompt:
Ask your LLM of choice to generate you a "literal, technical" prompt that is about 2 paragraphs long of which each paragraph has about 4 sentences that are written in a natural language style.
Latest version notes and model page updates:
major update, much MUCH better skin, with mostly keeping to the original FLUX style in the rest of the image (although a little bit is lost still)
LoRa description
This LoRa model was designed to improve the skin of people generated in a photorealistic style. It is not perfect. The skin is not perfectly real and it changes the image somewhat. It is still an improvement over the base, however.
Disclaimer:
I do not claim to be the current owner, original creator of, person or legal representative of a person depicted in, any of the concepts that my models, including this one, aim to emulate nor the training data used to train this model. I do not claim that any of my models have received an endorsement by those respective individuals. I also do not claim that the emulation attempted by my models represents a 100% accurate depiction of the original concept in question nor do I claim that the quality of my emulation reaches the same quality of the original concept. All credit goes to the respective current owners, original creators, or people and I encourage you to support them in any way you can.
If you are the current owner, original creator of, person or legal representative of a person depicted in, any of the concepts that any of my models aims to emulate and would like that specific model to be removed, then please leave me a private message here or on Reddit with proof of authenticity of your identity and claim and I will remove it.
Additionally, I do not endorse my models being used to violate the law or enact still legal, but immoral acts such as the spread of misinformation using deep fakes.
Description
FAQ
Comments (14)
I hardly ever leave comments, but that is a great job. well done
While i do think this is a great enhancement for realism, i do fear that the likeness change may be too strong. Ill do some testing as i have a video ill be showcasing 9b in with ASYM so ill include this lora as well
Whats ASYM?
you wrote "with skin with textured detail" as a trigger. maybe theres a "with" that need to be removed?
Please change the trigger word. writing textured skin with klein 9b makes weird imperfections like hairs or blood vessels on the body, so it would be cool to active u'r lora differently because for nsfw this is not usable
I have not seen any such things happening nor has anybody else ever reported such a thing. Also the trigger words arent randomly chosen. I cannot just retrain the lora with a different trigger word and everything works out perfectly on the first try. That stuff takes quite a bit of experimentation, which takes time and money.
@AI_Characters I was using 4 steps, and saw that you used 50. this is the only difference with you'r generations so I'll try and see if it helps, but 50 will take a lot of time, it's a lot.
Edit : completely burned with 50 steps. So what's wrong ?
SNOFS Lora 1.4, Character lora, you'r two loras, all weight : 1 - euler simple or euler ancestral beta, 1 cfg, 4 steps.
@winielourson2852 Youre probably using too many loras at once without reducing the strength. Also this is a base LoRA. not a distilled LoRA. the 50 steps is for base.
Same. I‘m not sure, but it seems like the trigger words are telling the base or turbo model to add random and unwanted details like pimples or birthmarks on its own and then applying the lora on top of it, which results in an overload of „details“. Changing the trigger words to something the base or turbo model aren’t recognising might fix this kind of behaviour in general…
@Iridu I have already tried a variety of trigger words. If I dont include textured or complexioned or such it does not get the skin texture right at all. It basically doesnt train. This is the best I can do at the moment.
What an amazing LoRa! Thank you so much for all your work on this.
One thing I've noticed when testing is that the strength settings seem to be the obverse of what I would have expected, with lower settings (like 0.6) having a much stronger effect. Was this by design, or is it because I'm using it with the distilled version of 9B (flux-2-klein-9b-fp8mixed) over four steps?
I dont know. I havent tried lower settings and I dont use distilled.
would you port you'r loras on krea2 ? it would be insane dude
Right now that seems unlikely. I am not able to get the same likeness results on Krea2. I am also very financially constrained these days. I will probably port some over but they will probably not be as good as my Klein versions and it probably won't be all of them.
Details
Files
FLUX.2-klein-base-9B_LoRa_by-AI_Characters_CONCEPT_BetterSkin_v2_TRIGGER$skin with textured detail$.safetensors
Mirrors
FLUX.2-klein-base-9B_LoRa_by-AI_Characters_CONCEPT_BetterSkin_v2_TRIGGER$skin with textured detail$.safetensors
FLUX.2-klein-base-9B_LoRa_by-AI_Characters_CONCEPT_BetterSkin_v2_TRIGGER$skin with textured detail$-mid_2613362-vid_2946217.safetensors
bttrskn2.safetensors
better-skin-v2.safetensors
FLUX.2-klein-base-9B-Better-Skin.safetensors
FLUX.2-klein-base-9B-Better-Skin.safetensors
FLUX.2-klein-base-9B_LoRa_by-AI_Characters_CONCEPT_BetterSkin_v2_TRIGGER$skin with textured detail$.safetensors
Flux2Klein9b_Better_Skin_v2.safetensors
Better Skin v2.0 - Klein9B - with skin with textured detail.safetensors
Better Skin v2.0 - Klein9B - with skin with textured detail.safetensors
skin.safetensors
better_skin_concept_v2.safetensors
FLUX.2-klein-base-9B_LoRa_by-AI_Characters_CONCEPT_BetterSkin_v2_TRIGGER$skin with textured detail$-mid_2613362-vid_2946217.safetensors
151-AI_Characters_CONCEPT_BetterSkin.safetensors
betterskin.safetensors



















