训练集收集自群友水无22，因此本模型命名为水无22。我也不知道这是具体什么画风
一个基于 Anima 训练的柔光系画风 LoRA，主要强化：
柔和自然的光影氛围
高曝光、低对比的梦幻观感
通透细腻的皮肤质感
淡彩色调与空气感表现
发丝透光与环境光融合
适合人物特写、海边、窗边、室内柔光等场景
整体风格偏向：
柔光、梦幻、空气感、奶油肌肤、浅色系、氛围感插画
触发词
shuiyue
推荐放在 Prompt 前部使用。
推荐参数
Base Model
Model: anima-base-v1.0
Model hash: bd43b7cffe
Module 1: qwen_3_06b_base
Module 2: qwen_image_vae
Generation
Steps: 36
Sampler: Euler a
Schedule type: Beta
CFG scale: 5
Shift: 3
Seed: 114514
Size: 1024x1536
Hires Fix
Denoising strength: 0.6
Hires upscale: 1.5
Hires upscaler: 4xHFA2k_ludvae_realplksr_dysample
Hires steps: 32
Hires CFG Scale: 5
Hires Shift: 3
Other
RNG: CPU
Emphasis: Original
Beta schedule alpha: 0.6
Beta schedule beta: 0.6
推荐权重
平衡模式
<lora:shuiwu:0.8>
兼顾风格与 Prompt 服从性。
推荐模式
<lora:shuiwu:1>
获得最完整的画风表现。
强风格模式
<lora:shuiwu:1.2>
风格影响更强，但可能降低部分 Prompt 自由度。
示例 Prompt
masterpiece, best quality, shuiyue, soft light, 1girl
推荐关键词
soft light,
dreamy atmosphere,
pastel colors,
window light,
backlighting,
glowing hair,
blurry background,
warm sunlight,
high exposure,
cinematic lighting,
delicate skin,
airy atmosphere
训练信息
Resolution: 1216x1664
Rank: 32
Alpha: 32
Learning Rate: 2e-5
Mixed Precision: bf16
Metadata
Lora hashes: "shuiwu-000045: b2965c7974b6"
Version: neo-2.24
感谢下载与使用，希望你喜欢这套柔光空气感画风。
Shuiwu Style LoRA
A soft-light anime illustration LoRA trained for Anima, focusing on dreamy lighting, delicate skin rendering, atmospheric color grading, and high-quality character illustrations.
Style Features
Soft and diffused lighting
High-exposure aesthetic
Smooth skin shading
Pastel color palette
Dreamy atmosphere
Delicate hair highlights
Gentle contrast
Strong ambient lighting
Cinematic mood
The overall style emphasizes:
Soft Light · Airy Atmosphere · Dreamy Colors · Delicate Skin · High-Key Lighting · Anime Illustration
Trigger Word
shuiyue
Recommended to be placed near the beginning of the prompt.
Recommended Settings
Base Model
Model: anima-base-v1.0
Model hash: bd43b7cffe
Module 1: qwen_3_06b_base
Module 2: qwen_image_vae
Generation
Steps: 36
Sampler: Euler a
Schedule type: Beta
CFG scale: 5
Shift: 3
Seed: 114514
Size: 1024x1536
Hires Fix
Denoising strength: 0.6
Hires upscale: 1.5
Hires upscaler: 4xHFA2k_ludvae_realplksr_dysample
Hires steps: 32
Hires CFG Scale: 5
Hires Shift: 3
Other Settings
RNG: CPU
Emphasis: Original
Beta schedule alpha: 0.6
Beta schedule beta: 0.6
Recommended LoRA Weight
Balanced
<lora:shuiwu:0.8>
Good balance between style strength and prompt flexibility.
Recommended
<lora:shuiwu:1>
Best overall visual quality and style representation.
Strong Style
<lora:shuiwu:1.2>
Maximum style impact with reduced prompt flexibility.
Example Prompt
masterpiece, best quality, shuiyue, soft light, 1girl
Suggested Keywords
soft light,
dreamy atmosphere,
pastel colors,
window light,
backlighting,
glowing hair,
warm sunlight,
blurry background,
cinematic lighting,
high exposure,
delicate skin,
airy atmosphere
Training Information
Resolution: 1216x1664
Rank: 32
Alpha: 32
Learning Rate: 2e-5
Mixed Precision: bf16
Metadata
Lora hashes: "shuiwu-000045: b2965c7974b6"
Version: neo-2.24
Credits
Model description and documentation generated with assistance from ChatGPT 5.5.