ImageEnhance2 / CharacterEnhance

While refining ImageEnhance, which improves images through conceptual manipulation, I separated the parts related to the whole picture from the parts related to the character. It is now possible to assign weights to each independently, which can help to some extent with the problem of grass growing when applied. Additionally, I lowered the LoRA network rank to reduce the impact on the image.

Original (Ishimura)

ImageEnhance2:1.0

CaracterEnhance: 1.0

Both

ImageEnhance is a tool for creating "better pictures" by enhancing concepts.

This Lora does not have image data. It reinforces the concepts built into the model that generally lead to good drawings. This Lora aims to bring out the maximum potential of the model.

Although it may affect a specific part, Lora is also related to overall elements such as composition, so even if you add it later, I cannot guarantee the composition before it was added. Additionally, some models may experience side effects such as drastic changes in their clothing.

Please start using around 0.5.

And, as the amount of application is increased, the screen tends to become darker. This may not be desirable depending on the purpose, so I will also release ImageEnhanceL to correct it. This adjustment is exactly the same, but a higher number is required to achieve the same amount of coverage.