>>> ride the wave <<<
>>> base model trained on past images and models
>>> simple prompt for great results,
>>> LoRA stacking on low weights give great results, examples in gallery,
>>> use negatives to change between realistic or 2D anime or illustration or flat or sketch,
>>> CFG scale 4.5 for normal use, and 1.1 or 1.2 for speed generations,
>>>
Description
trained with more images and results from version 1.0
FAQ
Comments (12)
With the Early Access Program, imma train the next LoRAs, :) THANK YOU FOR THE SUPPORT means a lot to me and PLEASE POST YOUR RESULTS !!!!
amazing work as always 🙏
Thank you !!!! i apreciate
I really like this checkpoint and use it a lot. It handle lots of different setting and subject. Great job. I have tons of images posted using this one. Try it out.
i apreciate, THANK YOU !!!!
best sdxl model , hope to get anima ver
its Illustrious, till Anima gets better i will stick with Illustrious
beautiful, i was on tensor and finding a "Leon Fauxlume XL | Illustrious 2.5D - v3.0" alternative to use on my pc and this is pretty much it, but without the 2.5 lora! and obiously putting that in its perfect, the detailing and the overall aesthetic is perfect even in low steps, its basically perfect!
THANK SO MUCH !!!! i apreciate, you can check out Okubo Dai v2.0 or v1.0 for solid 2D, and also n4mik4 v1.0 for a more solid 2D look, let me know, this comments are always welcome, dont forget to post some of your creations here :)
I've been obsessed with that model lately, and It hurts a lot to not be able to use it locally. I tried this one, yet I think it's still missing a lot, which Lora for 2.5d are you using?
@flimango, @Lunny15,
to be honest, this model doesn't really look like 'Leon Fauxlume'. The character anatomy is completely different — Fauxlume's is more aesthetic and refined. I did a whole investigation to figure out how the author of Fauxlume achieved that look. The model description says that 'Leon Fauxlume' is a merge of the OneObsessionV16 model plus some LoRAs that the author hand-picked. In the comments, the author doesn't reveal the LoRA secret. I ran many experiments, looked at examples of OneObsession on Civitai, searched for a similar style, wrote down and tested LoRAs, then went into Leon's image library on Tensor and checked what he used in his own generations. And I found an almost similar recipe. The body and skin texture are a 100% match — I'm still working on facial similarity, there are still some inaccuracies there. The author also used some LoRA to enhance the background, most likely 'Better Landscape', but it's paid, so I'll have to find an alternative. Same with color correction — that'll need some work. If you're interested, I'll post my results here.
@Ink0gnito I would be truly interested, thank you for your work!