Aemeath - Wuthering Waves outfit
General Information:
Best weight: 0.8. Recommend 0.6 ~ 1.0
Trigger word: aemeath
Appearance: 1girl, solo, long hair, very long hair, pink hair, ponytail, high side ponytail, ahoge, hair between eyes, hairclip, hair ornament, feather hair ornament, orange eyes, large breasts, halo, heart on chest, glowing heart
Outfit: white leotard, breastplate, elbow gloves, detached sleeves, bare shoulders, cleavage, gold trim, black choker, brooch, frilled skirt, thigh strap, single thighhigh, asymmetrical legwear, white boots
Description
FAQ
Comments (9)
Nice try, but still some small difference in the clothing, appearance, and seem to affect the camera angle and detail, hope it can be improved more in later iterations
I love your feedback! I'll take your points on board and work on them in the next version. Really appreciate you taking the time to test it out 👍️
The prompt for the eyes should be brown eyes, symbol shaped pupils, + +, yellow bottom iris, or you could shorten it a bit. Also, her appearance is too young, hope it can be aged up a bit to match in game...
Thanks for the eye tags, really helpful 😄
I think the younger look is mostly from the style LoRAs I used, especially Firedotinc. The character LoRA itself was trained on in-game screenshots and official 3D model images, so her age should be pretty close to the game. I'd recommend trying it with different style LoRAs and seeing which one fits her best.
@voltek it is still similar to my image on the model, I suggest anima base for best compatibility
Would you kindly train it for 2.9b anima? Thanks
I'd love to, but I'm training on an RTX 4060, and LoRA training on the 2.9B model is just too slow compared to the base model to be practical on my hardware right now. Sorry about that 🥲
Thank you for sharing! It's super useful. The replacement of clothes doesn't cause any pollution at all.
the lora is actually reproducing the character very faithfully, the best on the site. But unfortunately, it seem to be over trained or some setting/raw data problem, leading to style and posture bleeding into the generations.
Hope you can look into improving the lora, thanks!