These tag list(s) contain tags + their aliases (active & deleted + chars/artists/franchise tags) from Danbooru and/or e621 and are for use with sd-webui-tagcomplete extension and ComfyUI-Custom-Scripts autocomplete feature (or AutocompletePlus / comfy-ex-tagcomplete) as well as Krita AI Diffusion, Stability Matrix and SwarmUI (and lots more probably).
There are 3 main versions.
Danbooru - Danbooru tags only (best for Illustrious and NoobAI/Pony if not using e621 knowledge)
e621 - e621 tags only (best for e621 models/usage only; e.g: NoobAI)
danbooru_e621_merged - a merge of both Danbooru and e621 tags (best for NoobAI/Pony based models especially if you plan on using e621 stuff and your extension doesn't support loading multiple tag lists at once)
There's also šļø Krita AI Diffusion compatible tag lists inside the corresponding folder.
There is a šØš³ Chinese version made by @UrielGnoyzaki (comment).
The a1111-sd-webui-tagcomplete extension already comes with the 22.12.2024 tag lists so you don't need to download those from here.
22.12.2024 is the recommended version for overall use case if you don't want to switch a lot. "01.09.2025 ChenkinNoob 01" and similare are release made for the model of the same name, but not exclusive to, just use them for what you see fit. The ones that are have a more recent date are mostly there for future proofing and the older ones usually have worse filtering/merging but the site-indivdual tag lists should be ok to use.
You can find the full archive of all tag lists (including other misc stuff like some raw data) that have been made on the archive GitHub repository. (auto updates every months as GitHub actions artifacts, and uploads to repo every 3 months)
The script(s) with which I made these tag lists is open source on my GitHub.
Some info on the install process for some tools can be found on the main page of the repo (readme.md)
Some rather trivial info:
NoobAI 1.1/1.0 Eps/Vpred knowledge cutoff for danbooru is roughly before 23rd October 2024
NoobAI's e621 dataset is only up to date until early 2024 (Yi from Nine Sols will won't work for example) meaning older tag lists for e621 specifically might feel better to use
Illustrious v0.1/1.0-2.0 danbooru knowledge cutoff is roughly june 2024
RouWei has a slightly more up to date dataset (april 2025 for 0.8, dec 2024 for 0.7 or so) for those who use it and see this
Generally I don't think much has changed over the few months so the 22nd december 2024 will still be the recommended version.
The Danbooru tag scraping part was originally borrowed from BetaDoggo over here (many thanks)
Also a thank you to e621 for having every list easily accessible
also pandas dataframes are kinda cool.
Description
Recommended version for Mugen (š¤-link) - released on 31st March 2026; The model's knowledge is updated up to February/March of 2026. Visit their model description for more, important info.
Krita AI Diffusion compatible tag lists included.
Minimum post count threshold is 20
I'd like to suggest you to check out Mugen because I believe it to be a really cool project that is also worthwhile, and if you believe so too, please consider supporting them (be it for Mugen or other future endeavours).
IMPORTANT: Mugen may be based on Noob Vpred initially, but it did NOT include e621 dataset in their training and due to the significant change in the latents e621 tags may not work. Did not test, your mileage may vary.
FAQ
Comments (9)
Idk if there was a better way, but I had to scrape all of Gelbooru's tags instead of stopping at a certain threshold mid-fetch because I don't think I can sort by post count, so here is the full data. Maybe someone can make further use of it.
Btw it took 4-5 hours for full scrape, would've been 4-5x faster if Gelbooru didn't slow down the deeper I went
Somehow missing 'standing'?
@TyundrisĀ It is indeed missing standing in the CSV file, but fortunately not in the full data, so I can redo the CSV without hitting Gelbooru API
I'll look into it, I guess my CSV export is broken somewhere
ok i found out why, standing is marked as ambiguous tag, I felt like those should be removed but i guess not
If you redownload the zip file now it should have standing in the CSV
@DraconicDragonĀ How would standing be ambiguous? How would you prompt it otherwise? Upright on two legs??
Also, just my two cents, but truly ambiguous tags can be great wildcards. Thanks for looking into this :)
@TyundrisĀ Don't ask me how it would be ambiguous, ask Gelbooru's barely functioning system
The response from https://gelbooru.com/index.php?page=dapi&s=tag&q=index&json=1&names=standing&api_key=abcde12345&user_id=123_xyz is:
"tag": [ { "id": 1623, "name": "standing", "count": 1409136, "type": 0, "ambiguous": 1 ]
But also keep in mind that image board tags aren't made to be used for prompting, just tagging and organizing items/images
The tag is actually in use by Gelbooru, I guess maybe this is their "workaround" system for not having aliases; Maybe they also don't have implications either, though that is just an assumption since I haven't looked into that
@DraconicDragonĀ I'm not really bothered but head's up, apostrophes are turning into csv notation: '
@TyundrisĀ Thank you for catching that, other special characters also appear the same and it is also an issue in the full data...
And it's because Gelbooru's API returns them like this...
Anyway, I switched out the uploaded zip with a new one