V3:V0.1的升级，更好的平衡，颜色表现不俗。如果你想要有颜色，不要加monochrome, greyscale，演示图没有使用turbo lora，最后一版本(大概)。

V3:V0.1 upgrade, good balance, with impressive color performance. If you want color, don’t add monochrome or greyscale, showcases with out turbo lora, the last version(MAYBE?).

v2:target_res 1536 重训，更好更细腻的纹理。

v2: Retrained with target_res 1536. Better and more refined textures.

v1版本，更好的漫画表现力，色彩更少表现。

v1: better comic expressiveness, with less color expression.

基于base1.0训练的画风lora，T-LORA训练。主要用了漫画的素材训练，由于素材集里有码，所以相对难以消除，另外和ANIMA的彩色底模有时会在眼睛等地方展示出色彩，当然也基于这特性，你可以生成彩色的图片，提高monochrome 和 greyscale权重可以改善，实测3和4也能勉强可用。在训练时没有运用任何彩图，主要用在表现出漫画的感觉，可以配合多视角、分镜等词使用。

请试用。

触发词:@rmrstyle,

monochrome, greyscale（如果你想拥有色彩，不要加。）

如果想要更凌乱的线条，可以试着加入提示词:sketch,comic expressiveness

多视角提示词:multiple_views,frame,comic,comic expressiveness

LORA权重:0.8~1

Sampler: er_sde + beta

euler + normal

演示图:配搭了Anima Turbo LoRA v0.2 lora。

An art-style LoRA trained on base 1.0, trained with T-LORA. It was trained mainly on manga material; since the dataset contains mosaics (censorship), those are relatively hard to remove. Also, against ANIMA's color base model it sometimes shows color in places like the eyes — though by that same trait you can also generate color images. Raising the monochrome and greyscale weight improves this, and in testing 3 and 4 are barely usable. No color images were used at all during training, though — it's mainly meant to convey that comic feel, and can be used together with words like multi-view, frame, etc.

Please give it a try.

Trigger words: @rmrstyle,

monochrome, greyscale,(If you want color, don’t add)

If you want messier lines, try adding the prompt:sketch,comic expressiveness

Multi-view prompts: multiple_views,frame,comic,comic expressiveness

LoRA weight: 0.8–1

Sampler: er_sde + beta

euler + normal

showcase images: paired with Anima Turbo LoRA v0.2.