• 一种“较”真实系二次元风格，在经过一些神秘的lora配比后在k2上获得的风格（大嘘）

• 此风格聚焦于“较”真实的服装细节、皮肤以及良好的背景和光影表现，而不是类似3d的真实

• krea2版和moss版在魔搭上训练，krea2测试时使用的底膜为官方krea2 turbo int8

• 该风格画面较暗

• A "bit" realistic anime style, achieved after some mysterious LoRa matching

• This style focuses on a "bit" realistic clothing details, skin, as well as good background and lighting effects, rather than 3D-like realism

• The krea2 adn moss version is trained on ModelScope civision.The base model used for testing the krea2 version was the official krea2 turbo int8

• The style of the image is relatively dark

Moss(sft) v1.0

• sft版本在魔搭平台上训练，按moss模型的说明，sft版本训练出来的lora不适用于base版本 很快啊，一有魔搭能训练的风声就马上开炉😍

• 据魔搭平台的训练反馈，在此平台上训练的lora可能需要更新到最新comfyui版本才能被正确加载

• 后续会尝试训练base版本

• 感觉moss训练出来的lora比k2和anima的效果更显著

• 该版本可能存在部分细节模糊的问题

• The sft version is trained on the modelscope. According to the instructions of the Moss model, the lora trained by the sft version is not applicable to the base version

• According to the training feedback from the modelscope, the lora trained on this platform may need to be updated to the latest ComfyUI version to be loaded correctly

• I will attempt to train the base version in the future

• This version may have some issues with vague details

Anima v2.0

• 全新且更完美的画风，在听取一些人的意见后进行改进获得的最佳版本

• 更好的细节和背景，已经接近原数据90%的风格，剩下10%可能基于anima动漫画风的底层逻辑无法完全还原 我说k2大B就是好

• 用了188张素材训练（素材人物都有点偏幼），各细节都处理地非常好，但没测试过和别的lora进行配合

• 目前测试后凭感觉底膜anima-base、采样器er_sde配合simple效果最佳

• 没有触发词，即开即用😍

为了出好丹也是猛猛研究测试了一整个国庆，云端都花了我十几块，干了一堆炉渣

• The brand new and more perfect art style, the best version obtained after improvements based on some people's opinions

• With better details and background, it is close to 90% of the original data's style. The remaining 10% may not be fully restored due to the underlying logic of the anime style

• It was trained using 188 pieces of material, and every detail was handled exceptionally well. However, it has not been tested in conjunction with other LoRa devices

• Currently, based on our testing and subjective experience, the er_sde sampler paired with simple yields the best results.The recommended base model is anima-base

• No trigger word

Anima v1.0 Anima v0.9

• 画风较差，且为测试预览版，不建议再下载

• The artwork style is poor, and it is a test preview version. It is not recommended to download it

Krea2 v1.2

• 在优化了数据集和打标后，质感获得了进一步提升，但仍有部分模糊出现以及块状感

• 推荐搭配Detailer - beauty for Krea2

• After optimizing the dataset and labeling, the texture quality has been further improved, but there are still some blurring and blocky effects

• Recommended combination: Detailer - beauty for Krea2

Krea2 v1.1

• 优化了一下数据集，再进行更高一点的分辨率训练获得了1.1版本（

• 目前二采基本不会有太大的“硬化”效果，但可能会出现一些奇怪的细节和细节错乱

• 无论一采还是二采，在衣服褶皱细节方面已经接近原数据集80%的质感（感觉、应该、大概），但是头发依旧不太行，会有种3d方块的感觉，跟数据集中的柔和发丝还是差别很大。不过远距离视角下的人物头发确实是柔和的质感，近距离视角真跟方块一样还会部分细节错乱（不懂是不是过拟合）

• 目前正在尝试在现有lora上进行加训，如果效果还是不好会尝试再搞一份数据集重练，再不好就只能换训练网站或云端训练了（有一说一魔塔的训练参数少的可怜，打标也不能自己打，感觉目前问题大都出现在打标上）

• 推荐0.8的强度再加个1强度的Detailer - beauty for Krea2

• 后续有时间会考虑练个anima版本

• After optimizing the dataset and conducting training at a higher resolution, we obtained version 1.1

• Currently, secondary sampling generally does not have a significant "hardening" effect, but it may introduce some strange details and disruptions in details

• The texture of the hair is still not very good, giving a 3D blocky feel, which is quite different from the soft strands in the dataset. However, from a distant perspective, the character's hair does have a soft texture, but up close, it really looks like blocks with some details scrambled

• Currently, we are attempting to conduct additional training on the existing LoRa platform. If the results remain unsatisfactory, we will attempt to create a new dataset for retraining. If that fails, we will have to switch to a different training website or cloud-based training

• Recommend a strength of 0.8 and add a 1-strength Detailer - beauty for Krea2

• I'll consider practicing the anima version when I have time later

Krea2 v1.0

• 经测试后效果不是特别好，并且不适用于潜空间放大二采，即使是一采生成出来的图也有一些“硬化”的质感。当然，上述问题可能在后续的版本中进行改进

• 此风格可能会导致画面有种灰暗感

• 推荐增加“Anime in real”和“Detailer - beauty for Krea2”来获得更强的质感

• After testing, the effect is not particularly good, and it is not suitable for secondary mining of latent space amplification. Even the images generated from primary mining have a somewhat "hardened" texture. Of course, the aforementioned issues may be improved in subsequent versions,perhaps.

• This style may give the image a gloomy feel.

• It is recommended to add "Anime in real" and "Detailer - beauty for Krea2" to achieve a stronger sense of texture.