Origin

General tags:

Delete redundant ones.
Example: jewelry → keep only specific items like necklace or earrings. If there isn’t a more specific descriptive tag available, you can just stick with jewelry.
Delete tags like:
virtual youtuber, official alternate costume

Clothing details:

Example: change jacket → red jacket, and only keep the specific tag in the dataset.
swimsuit → keep only aqua bikini, frilled bikini and such.
Same goes for underwear as for swimsuit.
frills → frilled shirt or frilled piece of clothing (if a tag for that piece of clothing doesn’t exist, you can keep the general tag).
off shoulder → off-shoulder shirt or off-shoulder piece of clothing (if a tag for that piece of clothing doesn’t exist, you can keep the general tag).
school uniform & serafuku → these are a little bit tricky, you can keep them or remove them (I personally don’t use them if the dataset is simple).
pants or shorts → even if it’s not visible properly, stick to one which is correct (note: don’t forget to add color).

Pose / gaze corrections:

Remove “looking at viewer” if the character isn’t actually looking that way.
Ideally find a tag that describes gaze the best.
Don’t mix up “looking to the side” and “facing to the side,” as they are two different things.

Expression / Head / Body tags:

To get better control over face expressions, I recommend adding any missing tags like expressionless, shy, happy, etc.
Add closed mouth or open mouth in case it is missing.
Fix smile, frown, light smile.
Add colored eyelashes, but only where it’s visible.
Differentiate properly between fang, fangs, and fang out.
Stick to one eye color, e.g., aqua eyes or blue eyes, keep just one which is correct.
Same goes for hair color.
Fix short hair, medium hair, long hair, very long hair, etc…
⚠️ Careful: Close-up / portrait images cannot be combined with very long hair - Danbooru wiki will tell you more.
Unify breast size, and get rid of generic “breasts” tags.
Another tricky part is when it comes to breast tags:
- 98% of the time, I remove any breast tags where they are not visible.
- So basically, if there’s no visible cleavage or sideboob and such, I remove any breast tags.
- Some people do keep it and tag it everywhere, so this is up to you.

Missed tags: Autotaggers often skip:

from side, from behind, from above, from below → tag these manually.
Change upper body → close-up or portrait as appropriate.

Lighting / scene:

If you kept evening or sunset images, make sure they’re tagged, as autotaggers often miss them.
Add night to night shots.

Accessories consistency:

Neck ribbon, bow, or bowtie? Pick one tag per outfit and stick with it.
Is it hair ribbon, hair bow, hairclip? Prune generic tags like “hair ornament” if it’s already specific. You can keep it if there’s no specific tag for it.

Style tags (for screenshots):

Remove any style-specific tags such as “official style” or “anime coloring” on screencap images.
Tag each image with “anime screencap.”

Yes, I know the tag on booru is “anime screenshot.” However, this tag was changed in 2025 from “anime screencap” after illustrious was trained.

This should help make the LoRA more flexible with styles.

Background tags:

Add blurry background, not just “blurry” – auto-taggers miss these quite often.
Remove mistagged transparent background – white backgrounds usually get mistagged as such.
Fix sky → blue sky, etc.

Multiple characters on an image:

You have two options for this one:

Tag everything of both characters.
Remove any tags related to non-focused character and let 1boy/1other absorb it.

This is the option I use almost all the time, as I do not care about other characters on an image.

“The more you tag, the better you become at recognizing which tags are accurate, and over time the process gets faster—especially when you take advantage of Group Edits.”

🔎 How do you know if a LoRA is undertrained?

The character looks generic or not recognizable.
Outfits or hairstyles are inconsistent across generations.
Key features (like accessories, ribbons, or colors) are often missing or wrong.
Feels like the LoRA didn’t “lock in” the character yet.

🔥 How do you know if a LoRA is overcooked / burnt?

The model forces details everywhere (extra accessories, weird shading).
You start seeing artifacts or distortions (e.g., eyes melt, clothes blur, odd textures).
Weird hands appear more often than usual.
It loses flexibility—no matter what prompt you use, it spits out the same rigid pose/outfit/style.
Colors may look oversaturated or unnatural.

✅ How do you know if a LoRA is just right?

The character is consistently recognizable across multiple generations.
Outfits and hairstyles show up reliably but with some natural variation.
No weird artifacts, distortions, or “burnt” textures.
It plays nicely with prompts—you can still guide style, lighting, or composition without the LoRA overriding everything.

Bonus check:

Chinese Translated Version

通用标签：

删除冗余的标签。
示例：jewelry → 只保留具体的物品如 necklace 或 earrings。如果没有更具体的描述性标签可用，可以继续使用 jewelry。
删除以下标签：
virtual youtuber, official alternate costume
RepentStar Added: delete * breasts but reserve breasts it self.

衣物细节：

示例：将 jacket 改为 red jacket，并只在数据集中保留具体标签。
swimsuit → 只保留 aqua bikini、frilled bikini 等具体标签。
内衣标签同理。
frills → frilled shirt 或 frilled _衣服部件_（如果该衣物部件没有特定标签，可以保留通用标签）。
off shoulder → off-shoulder shirt 或 off-shoulder _衣服部件_（如果该衣物部件没有特定标签，可以保留通用标签）。
school uniform 和 serafuku → 这些有点棘手，可以选择保留或删除（如果数据集简单，我个人不使用这些标签）。
pants 或 shorts → 即使看不清楚，也要选择正确的标签（注意：别忘了加上颜色）。

姿势 / 视线修正：

如果角色并没有看向观众，则移除 “looking at viewer” 标签。
最好找到能准确描述视线的标签。
不要混淆 “looking to the side” 和 “facing to the side”，这是两件不同的事情。

表情 / 头部 / 身体标签：

为了更好地控制面部表情，我建议添加缺失的标签，例如 expressionless、shy、happy 等。
如果缺少相关标签，请添加 closed mouth 或 open mouth。
修正 smile、frown、light smile。
添加 colored eyelashes，但仅限可见的情况下。
准确区分 fang、fangs 和 fang out。
统一眼睛颜色，例如 aqua eyes 或 blue eyes，只保留一个正确的标签。
头发颜色同理。
修正短发、中发、长发、超长发等标签…
⚠️ 注意：特写 / 肖像图像不能与超长发结合使用 —— Danbooru wiki 会提供更多说明。
统一胸部大小，去除通用的 “breasts” 标签。 RepentStar Added: Maybe:|
另一个棘手的部分是胸部标签：
- 98% 的情况下，我会删除所有不可见的胸部标签。
- 基本上，如果没有明显的乳沟或侧乳等情况，我会删除所有胸部标签。
- 有些人会选择保留并在所有地方标记，所以这取决于你。
RepentStar Added: 我个人更倾向于删去头发长度/颜色

遗漏标签：自动标记工具经常跳过：

from side、from behind、from above、from below → 手动添加这些标签。
将 upper body 改为 close-up 或 portrait，视情况而定。

光照 / 场景：

如果保留了傍晚或日落的图像，请确保它们被正确标记，因为自动标记工具经常遗漏这些。
在夜景中添加 night 标签。

配饰一致性：

是 neck ribbon、bow 还是 bowtie？每套服装选择一个标签并坚持使用。
是 hair ribbon、hair bow 还是 hairclip？如果是具体的标签，可以修剪掉像 “hair ornament” 这样的通用标签。如果没有特定标签，可以保留它。

风格标签（适用于截图）：

删除 screencap 图像上的任何特定风格标签，例如 “official style” 或 “anime coloring”。
每张图像都标记为 “anime screencap”。

我知道 booru 上的标签是 “anime screenshot”。然而，这个标签在 2025 年从 “anime screencap” 更改，当时 illustrious 已经训练完成。

这应该有助于使 LoRA 在风格上更加灵活。

背景标签：

添加 blurry background，而不仅仅是 “blurry” —— 自动标记工具经常会遗漏这些。
删除误标记的透明背景 —— 白色背景通常会被误标为透明。
修正 sky → blue sky 等。

图像中的多个角色：

你有两种选择：

标记两个角色的所有内容。
删除与非焦点角色相关的所有标签，并让 1boy/1other 吸收这些内容。

这是我几乎每次都使用的选项，因为我并不关心图像中的其他角色。