This is my first Lora, and also the first training result.
这是我的第一个 Lora，也是第一个训练成果。
More suitable for 2.5D or even a bit more realistic painting style, vae recommends vae-ft-mse-840000-ema-pruned.safetensors
比较适配2.5D甚至现实一点的画风，vae推荐vae-ft-mse-840000-ema-pruned.safetensors
If you want to use anime style, it is recommended that vae use klF8Anime2VAE
如果想使用动漫风，建议vae使用klF8Anime2VAE
This is a lolita dress from my collection called Beautiful Secret, but since there are very few photos that can be trained on and they are all in dark field environments, the default will be on a dark background, and if you want to use a bright field, it is recommended that you use the (day:1.5)tag
这是我收藏的一条洛丽塔裙，名叫灿烂的馈赠，但是因为能够训练的照片很少，且都在暗场环境下，因此，默认会在黑暗的背景下，如果要使用明场，建议使用(day:1.5)tag
Description
增加了训练步数与训练图片投入量，适配了动漫风的模型，使用两种不同VAE进行测试，效果可自行观看，权重建议0.6-0.8之间
本次预览图还使用了Sharpness Tweaker LoRA(-1,1)，Saturation Tweaker LoRA(0.55,-0.55)，Brightness Tweaker LoRA(0.55,-0.55)，EnvyBetterHands LoCon(0.6)等进行调整画质。
Increased the number of training steps and the amount of training image input, adapted to the model of the anime style, the use of two different VAEs for testing, the effect can be viewed on their own, the weight is recommended between 0.6-0.8
This preview also uses Sharpness Tweaker LoRA(-1,1), Saturation Tweaker LoRA(0.55,-0.55), Brightness Tweaker LoRA(0.55,-0.55), EnvyBetterHands LoCon(0.6) and so on. Make adjustments to the image quality.
FAQ
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.