What Are Danbooru Tags and How to Use Them

What Are Danbooru Tags

Danbooru tags are a structured tagging system used to describe image content. They originate from the Danbooru image database and are widely used in AI image generation models (especially anime-based models like Illustrious, Anima, etc.).

Each tag represents a specific visual element:

• Character (e.g. hatsune_miku)
• Appearance (e.g. long_hair, blue_eyes)
• Clothing (e.g. school_uniform)
• Pose or action (e.g. looking_at_viewer, smile)
• Environment (e.g. outdoors, night)
• Style or quality (e.g. masterpiece, best_quality)

The goal is precision. Every tag adds information the model can interpret.

---

Why Danbooru Tags Matter

AI models trained on Danbooru-style datasets understand these tags very well.

Using correct tags:

• Improves accuracy of generated images
• Reduces randomness
• Gives you control over composition, style, and details

Using natural language instead (like full sentences) is less reliable for these models.

---

How Danbooru Tags Are Written

Danbooru tags follow strict formatting rules:

1. Lowercase only ✔ correct: blue_eyes
✘ wrong: Blue Eyes

2. Underscores instead of spaces ✔ long_hair
✘ long hair

3. Comma-separated list 1girl, long_hair, blue_eyes, smile

4. No unnecessary words ✔ 1girl
✘ a girl / one girl

Tags are keywords, not sentences.

---

Basic Tag Structure (Order Matters)

A solid prompt usually follows this structure:

1. Subject 1girl, 1boy

2. Character (optional) hatsune_miku

3. Appearance long_hair, blue_eyes, blush

4. Expression / Emotion smile, shy, serious

5. Pose / Action looking_at_viewer, sitting, walking

6. Clothing school_uniform, dress

7. Environment outdoors, night, city

8. Lighting / Atmosphere soft_lighting, glowing, depth_of_field

9. Quality Tags (optional) masterpiece, best_quality

Example Prompt (Correct Danbooru Style) 1girl, long_hair, black_hair, grey_eyes, blush, shy, looking_at_viewer, school_uniform, outdoors, cherry_blossoms, soft_lighting, depth_of_field, masterpiece, best_quality

How to Find Correct Tags

Use real Danbooru sources:

• Danbooru website
• Model training datasets
• Existing prompts from Civitai or similar platforms

Important: Not every word is a valid tag. If a tag doesn’t exist in Danbooru-style datasets, the model may ignore it.

---

Common Mistakes Beginners Make

1. Writing full sentences ✘ a girl with long hair standing outside
   ✔ 1girl, long_hair, outdoors

2. Mixing styles ✘ photorealistic + anime tags together (can confuse model)

3. Using invalid tags ✘ pretty_face (not always a real tag) ✔ detailed_face (valid in many models)

4. Overloading prompts Too many tags can reduce clarity. More is not always better.

Tips for Better Results

• Start simple → then add details
• Use real, known tags
• Group tags logically (subject → details → environment)
• Test and adjust (prompting is iterative)

---

Quick Beginner Template Use this as a base: 1girl, [hair], [eyes], [expression], [pose], [clothes], [environment], [lighting], masterpiece, best_quality

Example: 1girl, longblackhair, greyeyes, shy, lookingatviewer, schooluniform, outdoors, softlighting, masterpiece, bestquality

---

Key Takeaway Danbooru tagging is not about writing — it’s about describing images in structured keywords. If your tags are:

• Correct
• Clean
• Logical → your results will be significantly better.