Due to the insufficient training resources, the effects of excessive merging are very pronounced. Therefore, it is necessary to increase the prompt weights to reduce the impact. Personally, I find the visual presentation very good, but it is limited to specific scenes. In the download options, there is another version： Excessive merging ver It is the heavily merged version that I trained previously. It is not very sensitive to prompts and basically draws images according to the dataset. I uploaded it simply because I found it interesting, as it indeed matches the painting style 100%... but it is also 100% uncontrollable. Using the <a target="_blank" rel="ugc" href="https://civarchive.com/models/2560840/anima-turbo-lora?modelVersionId=2979642">Anima Turbo LoRA</a> can effectively repair some deformed artifacts and uncontrolled prompt elements. However, it has a significant impact on the painting style, so the weight must be carefully reduced. In certain scenes, even adjusting it to 0.1 can still dominate the style. It is a double-edged sword LoRA. Using Anima Detail Tweaker can effectively increase character details while improving the painting style. After use, there is a slight result resembling the original artist's "high-quality" painting, which is quite good. 🍌（For details, please refer to the nun example image）

Anima_fumihiko_style（ふみひこstyle）

1.High probability of excessive merging (artifacts/noise beyond the prompt words).2.Complex scenes have a 100% occurrence of artifacts/noise.3.Prone to incorrect limb expressions.4.Requires high-weight prompt words to influence the image output; otherwise, the image performance is very poor.

Files with SHA256: 364331ee5a79241c3a92b79c3f9d10edd16d59d9b593277da68a797f47f72e15