📌 Basic Info
All the training data comes from images I generated using a style I really loved that I rolled from gpt-image-2.
Train model: Krea2 base (Civit Onsite Train, only v1)
Weight: 0.6~1.0
best = 1.0
if you want to mix other loras = 0.6~0.8
Sampler + Scheduler: euler + beta
📝 Version Notes
v3✨
Remove some over fractalization detailed datasets from v2 datasets.
Now you can get a nice style without too many fractalization
👉 Pick this if you just want my favorite gpt-image-2's style and don't want to have so many detail on your pics
v2-exp (Experimental)
Trained on a brand-new dataset, freshly generated from scratch and completely different from v1.
This one is experimental.
I'm trying to get the sweat spot between over "fractalization" and clean detail.
Also fix the default girl is locked to woman rather then girl
👉 Pick this if you want a balanced, versatile version that keeps the style
v1
Same dataset from: Ps GPT-image-2 style v1 (anima)
Pros 👍: You get beautiful backgrounds without specifying details in your prompt, and no extra detailer needed.
👉 Pick this if you want more detail on your pics
