Hello ♥
for whatever reason you want to show me appreciation, you can: ❤️Ko-Fi❤️
Made around 5-6k Images with Flux the last 3 days and made this Lora with 50 (100 with v0.2) of my favorite images.
Will train way more in the next versions.
FLUX is way better than this lora (with text), but sometimes you get a cool result with good text.
Settings
0.5 - 1.0
start with 0.5
try with and without Trigger
Trigger
flux-style
careful, if you use V0.1 there is an other trigger to activate
Hope you get a lot of beautiful results ♥
Description
trained with 100 images
fixed the Trigger 🫠
fixed a bug with the lora name
fixed a training image duplicate
FAQ
Comments (44)
lets gooooooooooo
you can download models for comfyUI with the clips built in, meaning you an put them in you regular models and Flux around and find out
https://huggingface.co/marduk191/Flux.1_collection/tree/main
Thanks, I'll be using this to mimic a few FLUX gens XD.
Amazing work! Love it.
Thank you so much, glad you like it ♥ And thank you for the huge tip, wow! Hope we will se a lot of cool artworks from you :)
@AIDigitalMediaAgency I do realism so I saved it to test it using different models. I did one using your LORA and the results are amazing! So I will tweak it and work on it a little more. Thank you for creating it and the tip is just a fraction of what you deserve for doing such awesome work.
@VeronicaL thank you so much for your kind words ♥
v0.2 in the next 48 hours!
What about 0.3 :'D ?
@apo1s let me sleep ☠️
Yay 🤩 Can't wait to flex the flux style images 😉 iykyk
You are a genius, you saved me a lot of computer memory
haha thank you ♥
Wow, how many GPU's did you use to generate 6k Flux images in 3 days! I will give this a try but I'm obviously not expecting the prompt following of Flux to be magically transfer to SDXL.
just a 4090. 🫠 but the most images are with the small fp8 version.
Flux is way better than this lora! you are right
Imma mix this with midJourney Mimic and set my computer on fire
I am sad cause no Pony Flux ;'(
if you need help with prompting for Maintenance Contest, check out my Prompt Collection - https://civitai.com/articles/6557/prompt-collection-for-maintenance-contest
Good job, dude. The lora is pretty useful, because it massively improves hands in SDXL.
thank you so much, glad you like it ♥
Can confirm it improving feet too.
Please keep on training this. Looking forward to seeing how this advances.
hope can do Pony style nsfw with flux soon
That won't happen. Next Pony right now will be on AuraFlow due to the licensing.
ever thought of discovering a model and not trying to stick your dick into it?
@Narz no
@Omikonz this 'community' is 99% suburban male tweens, so this tracks
@Narz na… not really
It seems to be ignoring characters when I use it at the recommended weight of 0.5, the character can be seen at 0.3 weight.
if you can say what kind of character, then is can run some test or update for next version
@AIDigitalMediaAgency Like when I generate Wolf O Donell at 0.5 weight, there would be a different wolf along with a different style
For me, only mentioning the word "cat" does produce a cat head. I put "cat head" 1.5 in negative prompt and "cat ears" in positive prompt, but it remained. I think in one prompt "cat ears" 0.8 or lower worked. cat in negative prompt may work, but when I want both a cat and a human, this will no longer be possible.
same here
I am interested in the generated prompt words for your 6k images. How do you let comfyui automatically update the prompt words and then generate the images? Was a huge wildcard used?
I'm not sure if I understand your question correctly.
But with this workflow I generate new prompt variations from a single prompt:
thank you so much
Why is the base model SDXL and not FLUX? I do not understand???
This is a Lora for SDXL and it creates the FLUX Style :) This is not Flux
@AIDigitalMediaAgency Flux style? Like, it makes 1024x1024 images look like 8k?
@Lazman the other way around, it makes 1024x1024 looks like a real 1024x1024 image, and not a blurry mess.
@ElectricDreams Fair enough. SDXL does seem sadly lacking in detail. I typically have to upscale to at least 3k+ with remecri to get decent results.
Please consider renaming it to Flux-like Style and not "Flux lora for" because non-native english speakers are translating and misunderstanding
its a good point, thank you. changed the name :)
Very nice effects.
Details
Files
aidmaFluxStyleXL-v0.2.safetensors
Mirrors
aidmaFluxStyleXL-v0.2.safetensors
aidmaFluxStyleXL-v0.2.safetensors
aidmaFluxStyleXL-v0.2.safetensors
aidmaFluxStyleXL-v0.2.safetensors
aidmaFluxStyleXL-v0.2.safetensors
aidmaFluxStyleXL-v0.2.safetensors
aidmaFluxStyleXL-v0.2.safetensors
aidmaFluxStyleXL-v0.2.safetensors
aidmaFluxStyleXL-v0.2.safetensors
aidmaFluxStyleXL-v0.2.safetensors
aidmaFluxStyleXL-v0.2.safetensors
aidmaFluxStyleXL-v0.2.safetensors