Ancient style, above the gentle breeze - a FLUX version of small fresh ancient style.

0
0
0
0
AnimeAnime CharacterGirlStyle Boost
最近更新:首次發布:
Anime,Anime Character,Girl,LoRA,FLUX.1Image info
Anime,Anime Character,Girl,LoRA,FLUX.1Image info
Anime,Anime Character,Girl,LoRA,FLUX.1Image info
Anime,Anime Character,Girl,LoRA,FLUX.1Image info
Anime,Anime Character,Girl,LoRA,FLUX.1Image info
Anime,Anime Character,Girl,LoRA,FLUX.1Image info
Anime,Anime Character,Girl,LoRA,FLUX.1Image info
Anime,Anime Character,Girl,LoRA,FLUX.1Image info

Friends, due to its unique rules and ecosystem, the flux model is actually quite difficult to apply.

When uploading the model, I hope to share some principles of application.

First of all, flux has a unique guiding value, which requires the cfg value on the sampler to be 1 to take effect.

When using the XL model, I always set the CFG value to 5, which changes everyone's habits of setting parameters for the result image.

Increasing the CFG value can result in no image generated, or just noise.

Secondly, flux almost entirely follows the style of the training set for training, not optimizing the lines and strokes in a low model way like the XL model does. They each have their own advantages in this aspect. However, only a small part of the generated images are optimized and beautified.

Therefore, the author is currently in the exploration stage,

not as free as with the XL model.

Let's also give NaicowCat some time to progress,

I believe progress will be made soon!

討論

最熱門
|
最新
發送
即將推出
下載
(0.00KB)
詳細資訊
類型
線上生成次數
0
下載次數
0
推薦參數
推薦搭配Checkpoint
推薦的重量
1
CFG
3
VAE

圖庫

最熱門
|
最新