Files

236 lines
6.9 KiB
Markdown
Raw Permalink Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
# 提示词指南
本指南介绍如何为 Anima 模型编写高质量的提示词。
> **注意**:本指南主要针对使用 `generate_anima_image` 的 Danbooru 标签体系。如果你使用的是 `generate_qwen_image`,请参考 [QwenEdit 指南](QwenEdit-Guide)(QwenEdit 使用自然语言而非标签词缀)。
## 硬性规则
1. **画师必须带 `@` 前缀**:如 `@fkey, @jima`,否则几乎无效
2. **必须明确安全标签**:`safe` / `sensitive` / `nsfw` / `explicit`
3. **提示词不分行**:单行逗号连接,分行会影响效果
## 提示词结构
推荐的标签顺序:
```
质量 → 人数 → 角色 → 作品 → 画师 → 风格 → 外观 → 标签 → 环境
```
### 字段说明
| 字段 | 说明 | 示例 |
|------|------|------|
| `quality_meta_year_safe` | 质量、年份、安全标签 | `masterpiece, best quality, newest, year 2024, safe` |
| `count` | 人数 | `1girl`, `2girls`, `1boy`, `no humans` |
| `character` | 角色名 | `hatsune miku`, `serena (pokemon)` |
| `series` | 作品名 | `vocaloid`, `pokemon` |
| `artist` | 画师(必须带 @) | `@fkey, @jima` |
| `style` | 画风 | `anime illustration`, `watercolor` |
| `appearance` | 外观 | `long hair, blue eyes, twintails` |
| `tags` | 核心标签 | `upper body, smile, white dress` |
| `environment` | 环境/光影 | `sunset, backlight, outdoor` |
| `neg` | 负面提示词 | `worst quality, low quality, blurry` |
## 质量标签
### 正面(推荐组合)
```
masterpiece, best quality, highres, newest, year 2024
```
### 负面(基础模板)
```
worst quality, low quality, score_1, score_2, score_3, blurry, jpeg artifacts, bad anatomy, bad hands, bad feet, missing fingers, extra fingers, text, watermark, logo
```
## 安全标签
| 标签 | 说明 |
|------|------|
| `safe` | 全年龄向 |
| `sensitive` | 擦边/性感但不露骨 |
| `nsfw` | 成人内容 |
| `explicit` | 明确的成人内容 |
**重要**:必须在 `quality_meta_year_safe` 中明确指定其中一个。
## 画师推荐
### 稳定组合
- `@fkey, @jima` - 效果稳定,色彩明亮
### 热门画师
- `@wlop` - 精致写实风
- `@ciloranko` - 明亮可爱风
- `@ask_(askzy)` - 细腻唯美风
- `@nardack` - 柔和梦幻风
### 画师名格式规则
- 必须以 `@` 开头:`@fkey`,而非 `fkey`
- 名字中间**无下划线**:`@kawakami rokkaku`,而非 `@kawakami_rokkaku`
- 多画师用逗号分隔:`@fkey, @jima`(但稳定性下降,建议最多 1-2 位)
- 如果画师名本身含括号(如 `yd (orange maru)`),需要转义:`@yd \(orange maru\)`
- 不常见的画师可能效果不稳定
## 长宽比
| 类型 | 比例 | 适用场景 |
|------|------|----------|
| 横屏 | 21:9, 2:1, 16:9, 16:10, 5:3, 3:2, 4:3 | 风景、场景 |
| 方形 | 1:1 | 头像、图标 |
| 竖屏 | 3:4, 2:3, 3:5, 10:16, 9:16, 1:2, 9:21 | 人物立绘、手机壁纸 |
## 示例
### 基础示例
```json
{
"aspect_ratio": "3:4",
"quality_meta_year_safe": "masterpiece, best quality, newest, year 2024, safe",
"count": "1girl",
"artist": "@fkey, @jima",
"tags": "upper body, smile, white dress",
"neg": "worst quality, low quality, blurry, bad hands, nsfw"
}
```
### 角色示例
```json
{
"aspect_ratio": "9:16",
"quality_meta_year_safe": "masterpiece, best quality, highres, newest, year 2024, safe",
"count": "1girl",
"character": "hatsune miku",
"series": "vocaloid",
"appearance": "long twintails, aqua hair, aqua eyes",
"artist": "@fkey, @jima",
"style": "anime illustration, cinematic lighting",
"tags": "full body, stage, concert, singing, microphone, spotlight",
"environment": "night, neon, crowd silhouette, bokeh",
"neg": "worst quality, low quality, blurry, bad anatomy, bad hands, extra fingers, text, watermark, nsfw"
}
```
### 场景示例
```json
{
"aspect_ratio": "16:9",
"quality_meta_year_safe": "masterpiece, best quality, newest, year 2024, safe",
"count": "no humans",
"artist": "@fkey",
"style": "anime background, scenic",
"tags": "landscape, mountain, lake, reflection, clouds",
"environment": "sunset, golden hour, dramatic lighting",
"neg": "worst quality, low quality, blurry, text, watermark"
}
```
## LoRA(可选)
如果需要加载 LoRA,使用 `loras` 数组:
- `name`:LoRA 名称(可包含子目录)
- `weight`:LoRA 权重(默认 1.0,UNET-only)
### 技术细节
- LoRA 仅作用于 UNET(不影响 CLIP),在 UNETLoader → KSampler 之间链式注入 `LoraLoaderModelOnly`
- **`name` 必须与 `list_anima_models(model_type=loras)` 或 ComfyUI `GET /models/loras` 返回值逐字一致**,否则触发 ComfyUI 校验错误
- `list_anima_models(model_type=loras)` 仅返回存在同名 `.json` sidecar 元数据文件的 LoRA(强制要求)
- Windows 下 `/models/loras` 通常返回反斜杠路径(如 `_Anima\xxx.safetensors`),在 JSON 中写作 `_Anima\\xxx.safetensors`
- 本工具会自动规范化路径分隔符(`/` 或 `\` 都可以)
- 远程 ComfyUI 场景下 list 功能不可用(无法读取远程文件系统),但可直接传 `loras` 参数
一行命令获取 LoRA 列表(PowerShell):
```powershell
Invoke-RestMethod -Uri "http://127.0.0.1:8188/models/loras"
```
示例:
```json
{
"aspect_ratio": "3:4",
"quality_meta_year_safe": "masterpiece, best quality, newest, year 2024, safe",
"count": "1girl",
"artist": "@fkey",
"tags": "full body, fantasy outfit, glowing",
"loras": [
{
"name": "_Anima\\cosmic_kaguya_lokr_epoch4_comfyui.safetensors",
"weight": 0.9
}
],
"neg": "worst quality, low quality, blurry, bad anatomy, bad hands, extra fingers, missing fingers, text, watermark, logo, nsfw"
}
```
## 常见问题
### 手脚崩坏?
负面词加:
```
bad hands, extra fingers, missing fingers, bad feet, extra limbs
```
### 兽耳娘变异?
负面词加:
```
anthro, furry
```
### 构图不对?
- 使用 `upper body`、`full body`、`portrait` 等明确构图
- 1MP 分辨率下确保主体占画面比例足够大
### 画风不稳定?
- 使用推荐的画师组合
- 避免使用过于小众的画师
- 可以添加 `style` 字段辅助
---
## Reroll 与批量生成
### 重新生成(Reroll)
每次生成后会自动记录参数。可以基于历史重新生成:
- `reroll_anima_image(source="last")` — 相同参数,新随机 seed
- `reroll_anima_image(source="last", artist="@ciloranko")` — 换画师
- `reroll_anima_image(source="12")` — 回溯第 12 条记录
- `list_anima_history()` — 查看最近的生成记录
### 批量生成
- **`repeat`**(推荐):提交 N 次独立任务,每次随机 seed,显存与单张相同
- **`batch_size`**:单次任务内 latent batch,速度快但更吃显存
- 总张数 = `repeat` × `batch_size`
```json
{
"repeat": 3,
"aspect_ratio": "3:4",
"quality_meta_year_safe": "masterpiece, best quality, newest, year 2024, safe",
"count": "1girl",
"artist": "@fkey",
"tags": "upper body, smile",
"neg": "worst quality, low quality, blurry"
}
```