This Danbooru tagger looks at an image and writes out the tags Danbooru would use for it, in Danbooru's own format: tags separated by spaces, underscores inside each tag (long_hair, looking_at_viewer, school_uniform), and a rating tag such as rating:g. It is free, there is nothing to install, and you do not need an account on Danbooru or anywhere else.
People use a Danbooru tagger for three main jobs: working out what a particular tag is called, tagging their own artwork or image collections the way Danbooru-trained models expect, and preparing a tag list to paste into a Danbooru upload. This page is built for those jobs. If you only want a quick tag list for an AI art prompt, with spaces instead of underscores and brackets escaped for prompt boxes, the anime tagger does that. This page keeps the exact Danbooru format, and it checks every tag against Danbooru's own data.
Behind the scenes it is the WD14 tagger by SmilingWolf, often called the "WD tagger". It was trained on millions of Danbooru images tagged by the community, up to the end of 2022, and the tagging model runs right here in your browser.
How to use it
- Add an image. Drag it anywhere onto the page, click the box to choose a file, or copy an image and press Ctrl+V (Cmd+V on a Mac). You can drop several at once, or a whole folder. Each one appears straight away and fills in as it is analyzed.
- Wait a moment. The first time, your browser downloads the tagging model once and keeps it, so later pictures take a few seconds.
- Review the tags. Every tag is a chip with a grey number (how many Danbooru posts use it) and a percentage (how sure the tagger is). Click a chip to leave it out, or slide toward "Fewer, surer tags" for a shorter list.
- Copy. The text box under the chips is the finished tag string. You can type in it freely, then press Copy tags.
The Danbooru tag format
Danbooru separates tags with spaces, and uses an underscore wherever a tag name has a space in it. That is why a tag is written long_hair and never "long hair": a space would split it into two tags. A short result looks like this:
rating:g 1girl solo long_hair smile scarf snow outdoors looking_at_viewer
The options under the text box change the output:
| Option | What it does | When to change it |
|---|---|---|
| Use Danbooru's current tag names | Swaps tags Danbooru has renamed for their current names and leaves retired tags out (more on this below) | Turn it off to see exactly what the model predicted |
| Add rating tag | Adds rating:g, rating:s, rating:q or rating:e for general, sensitive, questionable or explicit | Turn it off if you only want descriptive tags |
| Include characters | Keeps recognized character names in the text | Off for a generic description that is not tied to one character |
| Commas between tags | Writes 1girl, solo, long_hair instead of 1girl solo long_hair | On for caption files and training tools that expect a comma-separated list |
The rating is the tagger's guess, learned from how Danbooru's community rated similar pictures. Check it yourself before you rely on it. Danbooru's own help describes the four levels in more detail, and the rating of a picture is a judgment call that sometimes differs between people.
Renamed and retired tags
The WD14 model learned from Danbooru as it was at the end of 2022, but Danbooru's tag system keeps changing. Two kinds of change matter here:
- Renamed (aliased) tags. When the community decides a tag has a better name, the old name becomes an alias: anything tagged with the old name is automatically given the new one.
- Retired (deprecated) tags. Some tags are no longer used at all, usually because they were too broad or too vague to be useful on their own.
A model that stopped learning in 2022 still predicts the old names. So we compared all 9,079 general and character tags the model can output with Danbooru's public tag data on October 15, 2026. Of those, 270 are now aliases of a different tag and 233 are deprecated. This page fixes the first group and leaves out the second automatically, and marks both on the chips so you can see what happened.
| The tagger says | Danbooru now uses |
|---|---|
baozi | steamed_bun |
gamepad | game_controller |
china_dress | qipao |
doughnut | donut |
beam_saber | energy_sword |
gojou_satoru | gojo_satoru |
The best-known retired tag is bangs. Danbooru's wiki explains that the plain tag would apply to almost every picture on the site, so it is no longer used, and posts are tagged with specific kinds of bangs instead. The model still predicts bangs on most pictures of people, because it learned it before the change. Others you may see marked as retired include looking_away and the broad colored footwear and headwear tags such as black_footwear.
If you are training or captioning for a model that was itself trained on old Danbooru data, the old names may be exactly what you want. That is why "Use Danbooru's current tag names" is a switch, not a rule.
What the grey numbers mean
The small grey number on each chip is how many posts on Danbooru use that tag, as of October 15, 2026. It is a quick way to judge a tag:
- Big numbers mean a well-established tag.
1girlis on about 8.5 million posts,soloon 7.2 million andlong_hairon 6.3 million. Both the tagger and most Danbooru-trained generators know these tags very well. - Small numbers mean a niche tag. Characters are far rarer than the common tags.
hatsune_miku, one of the most popular, is on about 148,000 posts, and half of the 2,132 characters the model knows have fewer than about 2,200. A prediction for a niche tag was learned from fewer examples, so look at it twice. - Counts help you choose between near-synonyms. When two tags describe the same thing, the one with far more posts is the one the community actually uses.
The counts come from Danbooru's public tag data, saved on October 15, 2026. They keep growing, so treat them as a guide to how common a tag is, not as an exact live figure.
What this tagger gives you, and what it doesn't
Danbooru sorts every tag into one of five types, shown in different colors on the site. The tagger covers some of them well and some not at all:
| Type | Examples | From this tagger? |
|---|---|---|
| General (blue) | long_hair, smile, school_uniform | Yes. This is almost everything it outputs. |
| Character (green) | hatsune_miku | Yes, for 2,132 popular characters, and only when it is at least 85% sure |
| Copyright (purple) | vocaloid, original | No. A character tag often carries its series in brackets, as in ganyu_(genshin_impact), but you add the series tag yourself. |
| Artist (red) | the name of whoever drew it | No. A picture does not tell the model who made it. |
| Meta (yellow) | highres, commentary, translated | No. These describe the file or the post, not what is drawn. |
Because of the gaps, treat the output as a strong first draft. Add the artist, the series and a source yourself if you are going to publish it anywhere that expects them.
Tagging the Danbooru way
Danbooru has a few habits that are worth copying even when you are tagging a private folder, because every Danbooru-trained model learned from tags written this way:
- Tag what you see. Danbooru's guideline is to tag what is in the picture, not what you know about it. If a character is not shown, the tag does not belong.
- Avoid opinions. Subjective words such as "cute", "sexy" or "beautiful" are not tags. Describe the features instead:
smile,blush,long_hair. - Use the most specific tag. Danbooru links related tags, so a specific tag automatically brings in its more general ones.
very_long_hairimplieslong_hair, andcat_earsimpliesanimal_ears. You do not need to add both, because Danbooru adds the more general tag for you. - Count the characters. Tags such as
1girl,2girlsandmultiple_girlssay how many people there are, and they are among the most-used tags on the site.
Our guide to Danbooru tags goes through the system in more depth, and the tag glossary lists the common tags and what they mean.
Using the tags for a Danbooru upload
You can paste the tag string into the tag box when you upload a picture to Danbooru. A few honest points before you do:
- This site is not affiliated with Danbooru. It is an independent tool that uses Danbooru's public tag data.
- Add what the tagger can't. Danbooru expects an artist tag, the series (or
original) and a source for new uploads. The tagger can't see those, so fill them in yourself. - Check the rules first. Danbooru's upload rules do not allow fully AI-generated artwork. This tool is for tagging pictures you are entitled to tag, such as your own datasets or your own drawings, and for finding tag names. It is not a way to post generated images there. Read Danbooru's current rules before you upload anything.
Other boorus and training datasets
Many image boards, such as Gelbooru and Safebooru, use similar conventions: lowercase tags, underscores for spaces and tags for what is visible. Each site has its own rules and its own set of tags, so a tag that exists on Danbooru may not exist elsewhere and the other way around. Check the site's help before assuming the same names will work.
For machine learning, the Danbooru format is the one most anime image models were trained on. If you are tagging a whole dataset for a LoRA or similar, the batch tagger processes a folder at once and saves a caption file for every image. The LoRA captioning guide explains how to clean up and structure those captions, and tag pruning covers which tags to remove.
Where it makes mistakes
- It is trained on anime and illustration. It will describe a photo with tags such as
outdoorsandsky, but its vocabulary is built for drawings. - Tiny details get lost. Every picture is shrunk to 448 by 448 pixels first. Crop to the part you care about if it is small in the frame.
- Its knowledge stops at the end of 2022. Characters, outfits and trends that appeared afterward are not in its vocabulary, so it may describe them with older, more generic tags.
- Percentages are not proof. A tag at 90% is usually right and one at 40% is often wrong. The slider moves the cut-off; our guide to confidence thresholds explains how to choose it.
Frequently asked questions
What is a Danbooru tagger?
It is a tool that looks at a picture and suggests the tags it would have on Danbooru, an image board where every post is labeled with short tags by volunteers. AI taggers such as this one learned from millions of those labeled posts, so they can tag a new picture in the same style.
Is this Danbooru's own tagger?
No. This site is not affiliated with Danbooru. It uses the WD14 tagger model, which was trained on Danbooru data, and Danbooru's public tag information for the post counts and for the renamed and retired tag checks.
Why are some tags marked "renamed" or "retired"?
Danbooru keeps cleaning up its tags, and the tagging model stopped learning at the end of 2022. A "renamed" tag is one Danbooru now writes differently, so the page swaps in the new name. A "retired" tag is one Danbooru no longer uses, so the page leaves it out of your text. Untick "Use Danbooru's current tag names" if you want the model's original output.
How do I paste the tags into a Danbooru upload?
Press Copy tags and paste the text into the tag box. Danbooru expects tags separated by spaces, which is how this page writes them by default. Add the artist, the series and the source yourself first, because the tagger cannot see those. Check Danbooru's upload rules too: they do not allow fully AI-generated artwork.
What is the difference between this page and the anime tagger?
They use the same model. This page writes the tags in Danbooru's own format, shows each tag's post count and fixes renamed and retired tags. The anime tagger formats the same tags for AI art prompts, with spaces instead of underscores and brackets escaped for prompt boxes.
Where do the post counts come from?
They come from Danbooru's public tag data, saved on October 15, 2026. Counts keep rising as people upload, so the numbers show how common a tag is rather than a live total.
Can I tag a whole folder?
You can drop up to 100 pictures on this page at once. For bigger sets, or to save a caption file for every image, use the batch tagger. It runs the same model and can export one .txt file per image.