t8mars/comfyui-anima-t8

comfyui-anima-t8 is an add-on for ComfyUI, an open-source node-based interface that people use to drive image-generation AI models. This particular add-on is built around an anime-focused model called Anima, and its job is to help users assemble the long text prompts that the model needs in order to produce a picture. The add-on adds four nodes to the ComfyUI canvas: one that combines a positive prompt, a negative prompt, and a style prompt into a single piece, one that picks an artist style and also shows preview images, one that joins two text pieces, and one that loads a previously saved prompt by ID. On top of the nodes, it injects extra panels into the ComfyUI main page so users can browse a style library of 52 ready-made presets in 15 categories, a gallery of more than 1000 artists, and Danbooru tag lists covering artists, work names (such as anime or game titles), character names, and stylistic meta tags. A Civitai button is also provided. The user types in the numeric ID of a model on civitai.com, and the add-on calls the public Civitai images API to pull the highest-rated example pictures, reads the prompts attached to each image, sorts the tags into 14 internal categories, and saves the top 30 of them into the local style library so they can be reused later. Installation is done either by cloning the repository into ComfyUI's custom_nodes folder and running pip install -r requirements.txt, or through the ComfyUI Manager once the project is listed there. The dependencies are Python 3.10 or newer, a recent ComfyUI build, plus Pillow, numpy, torch (already included with ComfyUI) and requests. Under the hood, all fetched data is cached to a local SQLite file so panels open quickly on the second visit, and Danbooru images are routed through a same-origin proxy inside the ComfyUI server to get around anti-hotlink and content-security blocks. The README also lists the HTTP routes the add-on registers and a detailed version history covering recent fixes to preview rendering and tag-name handling.