🎨 59k+ Danbooru Artist List (Anima 2B Training Snapshot)
This Index contains 59,676 Danbooru artist tags recognized within the Anima 2B model.
Link: https://github.com/ThetaCursed/Anima-Style-Explorer/blob/main/Anima2B_Artist_Index_59k.txt
⚠️ WHY THIS LIST IS CRITICAL:
Do not rely on live Danbooru tags. Since the model's September 2025 knowledge cut-off, many artists have changed aliases, deleted accounts, or moved platforms.
Static Metadata: This list represents the exact "frozen" state of the training data.
Guaranteed Recognition: If a tag is in this file, it is part of the model's internal weights. Using these specific strings ensures the model actually "understands" the requested style.
📋 TECHNICAL DETAILS:
Frequency Sorted: Artists at the top have the most training samples. Artists at the bottom have the fewest.
Format: Raw
.txtfor easy searching and script integration.
⚖️ DISCLAIMER:
Independent Discovery: This list was compiled independently through data analysis.
No Affiliation: I have no contact with the creators of Anima 2B. This is a third-party resource provided for the community to improve prompting accuracy and explore the model’s stylistic range.
This doesn't seem to be entirely accurate.
For example, the tag "@bunchi" isn't known by anima, but is on this list. But the tag "@otintin" (which was the previous tag for this artist) is known by anima, but isn't on this list.
Why "@bunchi" is in the list (and why the model knows @otintin)
The list is correct based on the developers' statement that the knowledge cut-off is September 2025.
Here is why there is a discrepancy:
- The Timeline: The artist changed their tag from
"@otintin"to"@bunchi"in April 2025. Since this happened before the September cut-off,"@bunchi"is the correct, up-to-date tag in the metadata at that time. - The Model's "Memory": The artist spent over a year (March 2024 – April 2025) posting as
"@otintin". The model saw hundreds of posts under the old name, but only a few months' worth under the new one.
Conclusion: My list correctly reflects the database status as of September 2025. The model simply recognizes the old tag better because it has a much larger "weight" in the training data from the previous year.
Verification (Danbooru):
Shouldn't you change 'Guaranteed Recognition: If a tag is in this file, it is part of the model's internal weights. Using these specific strings ensures the model actually "understands" the requested style.' then?
Whether the cut-off knowledge is correct or not doesn't really matter if the tags in your list don't actually correspond to the style of that artist.
Rare exceptions, such as the artist tag you provided, have little impact, as the other artist names are correct.
Furthermore, once the model completes their training, this list will become even more relevant and accurate.