I found that images generated with my LoRA appear bit more amateur and less like the puppy "model_flux" style.
In Better Faces, I aimed to include a diverse range of cultures and ethnicities—15 if I remember correctly—with both young and old subjects, currently focusing on women.
aborigines, african, arab, arctic, brazilian, chinese, egyptian, finish, german, havaiian, indian, japanese, mongolian, russian, western
my FLUX-lora responds to age prompts like __yo for 20yo, 30yo, 40yo.
Weight: start with 0.3
look image post for more, i was curious also negativ has an impact^^
V........... see below ..............V
Try my other face expression LoRA. If you use it at a weight of only 0.4 (without trigger words), you'll achieve more realistic results.
https://civarchive.com/models/151594?modelVersionId=827638
I aim for realistic faces with natural skin, avoiding perfect symmetry. :)
FLUX with Euler-Sampler and SGM-Uniform-Scheduler performs well at -> 15-25 Steps
Trigger words aren't essential but are useful for specifying cultural features! and "yo" ;)
Model Descriptions:
FLUXV3: new image set, new caption, 15 cultures, no permission to merge!
FLUX: Recognizes mature (40yo) and young (20yo) subjects with reduced bokeh effect. For higher weights, it learns faces without jewelry, so consider adding jewelry and background details if desired.
pony_weak: Better for close-ups; doesn't perform well with complex prompts. Not limited to "1girl" and includes backgrounds.
pony_strong: works better with more complex prompts, not only 1girl, and call background
face enhancer_v3: Stronger in close-ups. Please consider using my prompts. ;)
face enhancer_v2: Much smoother and improved.
Multi-Face Models:
multifacesv_3: (Keywords: group, crowd, many faces) Generates more people, but distant faces still need improvement. Recommended weight: 0.3-0.8.
multifaces_v2: (Trigger: specify the "number" of faces) i used also copax and juggernaut that is already not that bad at small faces so may you dont need it!
multifaces_v1 -> (trigger: group, faces)
Additional Recommendations:
Overall, I suggest using Adetailer or a refiner for enhancing small faces.
Face Swap with Adetailer and Reactor
https://civarchive.com/articles/5002/face-swap-with-adetailer-and-reactor
The new SDXL VAE (released on 2023.07.26) is significantly better than older versions for face generation.
Please try my LoRA, and you'll often get more realistic faces—not those blurred, soft ones. ;)
Hints:
please, almost no negative prompt is necessary!
see my articles:
https://civarchive.com/articles/1890/better-faces-lora
https://civarchive.com/articles/1826/sdxl-prompts
All SFW!
Description
it doesn't really make the faces better or more detailed, but it knows mature and young and reduce the SHITTy BOKEH
start slowly at 0.3
FAQ
Comments (8)
One of my favorite things to do in the pre-Pony days was the generate interesting mixes of women. I'd have wildcards set up listing dozens of nationalities and had so much fun seeing the variations it created. Pony and pony-derived models seemed to really only understand Caucasian and Asian ethnicities with any reliability, so having such a LORA is really neat.
THX its not that superb ... but i dont like these puppy faces ^^
God thinks your description should be rewritten in a clearer way. Especially if you are asking people to pay to use your lora. God created languages and he still has a hard time understanding.
yes iam a non-native speaker ... and fyi one training (and its not allways good at first shot) cost 2000cred
@sevenof9247 AI Chatbots sind Ihr Freund, mein Freund. Einige sind besser als andere, aber die meisten, wenn nicht sogar alle, würden Ihnen helfen. DeepL ist ziemlich gut, aber das können Sie selbst beurteilen, da dies von ihm geschrieben wurde und ich kein Wort Deutsch spreche. Wie auch immer, vielen Dank für Ihre Bemühungen und Modelle!
I second this. It's not about sounding native. I'm also a non native speaker. It's about being intelligible. This is very hard to read. I gave the model description to OpenAI o1 with a prompt to make it intelligible and it produced this. (You may want to edit your description) It's much easier to understand:
__
Left Image: >>> With my LoRA <<<
Right Image: >>> Without (the puppy face) <<<
---
I found that images generated with my LoRA appear more amateur and less like the "model_flux" style.
In Better Faces, I aimed to include a diverse range of cultures and ethnicities—11 if I remember correctly—with both young and old subjects, currently focusing on women.
It responds to age prompts like __yo for 20yo, 30yo, 40yo.
---
Try my other face expression LoRA. If you use it at a weight of only 0.4 (without trigger words), you'll achieve more realistic results.
Link to the model
Be cautious: it enhances images and requires fewer steps (around 12 for portraits), so feel free to experiment.
I aim for realistic faces with natural skin, avoiding perfect symmetry. :)
Attention: Using a weight between 0.3 and 0.9 with 25 steps on DPM++ 2M SDE Karras yields images almost identical to those generated without it in 70 steps.
FLUX with Euler and DDIM samplers performs well at 12-20 steps.
Trigger words aren't essential but are useful for specifying cultural features! ;)
---
Model Descriptions:
FLUX: Recognizes mature (40yo) and young (20yo) subjects with reduced bokeh effect. For higher weights, it learns faces without jewelry, so consider adding jewelry and background details if desired.
pony_weak: Better for close-ups; doesn't perform well with complex prompts. Not limited to "1girl" and includes backgrounds.
pony_strong: Handles more complex prompts effectively. Not limited to "1girl" and includes backgrounds.
face_enhancer_v3: Stronger in close-ups. Please consider using my prompts. ;)
face_enhancer_v2: Much smoother and improved.
---
Multi-Face Models:
multifaces_v3: (Keywords: group, crowd, many faces)
Generates more people, but distant faces still need improvement.
Recommended weight: 0.3-0.8.
multifaces_v2: (Trigger: specify the "number" of faces)
Designed for the base model; feel free to try or merge it to see the results.
I also used Copax and Juggernaut models, which handle small faces well, so you might not need this.
multifaces_v1: (Trigger: group, faces)
Combine with v3 for better results and let me know your feedback. ;)
---
Additional Recommendations:
Overall, I suggest using Adetailer or a refiner for enhancing small faces.
Face Swap with Adetailer and Reactor
The new SDXL VAE (released on 2023.07.26) is significantly better than older versions for face generation.
But try my LoRA, and you'll often get more realistic faces—not those blurred, soft ones. ;)
@saldiamneo thx a lot ... I'm also too impatient to describe it all ^^
@jrewingwannabe947 I don't know if you even care, but your DeepL text in German is pretty good. Grammar and punctuation are excellent - that's where even many humans struggle. However, as a native speaker, I still get a slight AI vibe from the wording.
(This reply has been written by myself - all shortcomings are completely my own.)