Anima Gelbooru Style T8

Anima Gelbooru Style T8 is the pack's newer sibling to Anima Artist Style T8, and it does the same job against a different tag source: Gelbooru instead of Danbooru. If you've spent time on both boards you already know why that's a real distinction - same vocabulary family, but Gelbooru has its own artist, copyright, character, and general tags, some pools and conventions that only exist there. It arrived in v1.5.0 of comfyui-anima-t8, so it's the youngest node in the pack and a touch rougher around the edges than the Danbooru one, but the shape is identical.

How it works. One required textarea, gelbooru_tags, with comma- or newline-separated tags. The @ marker means "this is an artist" (so Anima treats it as a style trigger rather than a random tag), and any tag can carry a :weight. default_weight (0.1–2.0) covers anything you left unweighted; use_artist_prefix toggles whether the @ prefix survives into the output. From that, build() produces the STYLE_PROMPT string - @wlop:1.1, 1girl, long_hair style output - and a PREVIEW_IMAGES tensor with representative thumbnails for the tags you picked, pulled from Gelbooru through the pack's local proxy.

The optional last_picked input works exactly like the artist node's: the gallery's "add selected" button writes your current picks there, PREVIEW_IMAGES shows only those, while STYLE_PROMPT still includes everything in gelbooru_tags. Clear it and previews fall back to the full list.

Where it differs from the Danbooru node. The tag library button above the node opens four Gelbooru tabs - artist, copyright, character, general - versus the Danbooru node's five (which adds the style-meta and mooshieblob tabs). If you only care about artist style, the Danbooru node is the deeper well: 30,000+ artists versus what Gelbooru offers. The Gelbooru node earns its place through the general-tag panel, which is handy for browsing broad concepts Gelbooru happens to tag more carefully. It also has a documented resilience quirk: Gelbooru's official API returns 401 without a key, so the client falls back to the public autocomplete and HTML thumbnail endpoints automatically. You don't need an API key, and that fallback is why the tag search and previews still work out of the box.

Wiring and install. Same recipe as the artist node: STYLE_PROMPT into Anima Prompt T8's style field or a CLIPTextEncode, PREVIEW_IMAGES into a PreviewImage node. Installation is the pack-wide one - ComfyUI Manager search comfyui-anima-t8, or:

cd ComfyUI/custom_nodes
git clone https://github.com/T8mars/comfyui-anima-t8
cd comfyui-anima-t8
pip install -r requirements.txt

then restart. requests is the only added dependency.

The gotchas are the shared family ones: previews need internet and route through your local ComfyUI server's proxy, tags are cached in SQLite after first browse, and a tag with no results renders as a black placeholder tile. If you're choosing between the two style nodes, think of it as a tag-source preference, not an upgrade - pick the board you actually browse, and grab the other node if you want both libraries on the canvas at once.