This anime tagger looks at one picture and lists what it sees, using the same short tags the Danbooru community uses: 1girl, long_hair, smile, school_uniform, cherry_blossoms and thousands more. You can copy the result straight into an AI art prompt, use it to caption a picture, or just find out what the tags for a style or outfit are called.
Behind the scenes it is an AI tagger called WD14 (often shortened to "WD tagger"), trained on millions of tagged Danbooru images. It runs right here in your browser instead of on our servers, which is why nothing gets uploaded.
How to use it
- Add a picture. Drag it onto the box above, click to choose a file, or copy an image and press Ctrl+V (Cmd+V on a Mac).
- Wait for the tags. The very first time, your browser downloads the tagging model, about 388 MB. After that it is saved in your browser, so the next picture takes a few seconds.
- Tidy up and copy. Click any tag to leave it out, move the slider for more or fewer tags, then press Copy tags.
What you get back
Results come in three parts:
- Tags: everything the tagger is reasonably sure it can see, most confident first. The percentage on each tag is how sure it is.
- Characters: named characters it recognizes, such as
hatsune_miku. It only lists a character when it is very sure (85% or more), because a wrong name is worse than none. The character tagger page focuses on this part. - Content rating: one of Danbooru's four ratings (general, sensitive, questionable or explicit). It's shown under the picture and left out of your copied tags unless you tick "Include rating".
For a typical illustration of a girl in a winter scene you might see something like 1girl, solo, scarf, snow, outdoors, long hair, smile, coat, winter, looking at viewer. New to these tags? Danbooru tags, explained covers what they mean, and the tag glossary lists the common ones.
Using the tags as an AI art prompt
Many anime-style image generators were trained on Danbooru tags, so a tag list makes a good starting prompt. A few switches under the results get the text ready to paste:
| Option | What it does | When to use it |
|---|---|---|
| Spaces instead of underscores | Writes long hair instead of long_hair | Most prompt boxes. Turn it off if your tool expects the original underscores. |
| Prompt-ready brackets | Writes ganyu \(genshin impact\) instead of ganyu (genshin impact) | Interfaces such as AUTOMATIC1111 and Forge, where plain brackets change how strongly a word counts |
| Include characters | Keeps recognized character names in the text | On for prompts about that character, off for a generic look |
| Include rating | Adds the content rating tag | Some anime models use rating tags as a prompt switch |
Before you paste, it helps to remove tags that describe things you don't want to repeat, such as a watermark or signature, and to move the most important ideas to the front. Prompt length matters too: every extra tag dilutes the others a little, so a focused list of the tags you actually care about usually works better than every tag the tagger found.
Getting better results
- Use a clear, reasonably large picture. The tagger shrinks every image to 448 × 448 pixels, so tiny details in a big scene can get lost. If you care about one character in a crowd, crop to them first.
- Use the slider. Slide toward "More tags" to catch small details (with a few more mistakes), or toward "Fewer, surer tags" for a short, reliable list. The default of 35% is a good middle ground. Our guide to confidence thresholds explains the trade-off.
- Expect anime and illustration to work best. The tagger learned from drawings. It will still describe a photo (
outdoors,sky,cat), but its vocabulary is built for art. - Transparent backgrounds turn white. The picture is placed on white before tagging, so a cut-out character often gets
white_backgroundorsimple_background.
Tagging lots of pictures at once
This page tags one picture at a time, which is perfect for prompts and quick questions. If you're captioning a training set or organizing a folder, use the batch WD14 tagger instead. It tags hundreds of images in one go, lets you search and bulk-edit the tags, and downloads a .txt caption file for each image. It uses the same model, so if you've already used this page, the model is already in your browser and won't download again. Our LoRA captioning guide walks through the whole process.
Frequently asked questions
What is a Danbooru tagger?
It is a tool that looks at an image and suggests the tags it would have on Danbooru, a large anime art site where every picture is labeled with short tags by volunteers. AI taggers like this one learned those tags from millions of labeled examples, so they can tag new pictures the same way.
Is it really free and private?
Yes. There is no account, no limit and no watermark. The tagging happens on your own computer, so your picture is never uploaded. The page shows ads, which is how the site pays for itself. Our privacy policy has the details.
Can it tell me which character is in the picture?
Often, yes. It recognizes 2,132 popular characters and lists them under "Characters" when it is confident. The character tagger puts that front and center, shows close guesses, and lets you search the full list.
Is this the same as the WD tagger or WD14 tagger?
Yes, it uses the WD14 ConvNeXt V2 tagger by SmilingWolf, the model many people call the WD tagger. The WD14 model comparison explains how it differs from the other versions you can pick in the batch tagger.
Why does it work better on a computer than a phone?
The model is a 388 MB download and needs a fair amount of memory to run. Most laptops and desktops handle it easily; some phones run out of memory or take a long time. If it fails on your phone, try a computer.