shiroeirin/comfyui-good-anima

ComfyUI Good Anima is a set of AI agent skills for generating anime style images using the Anima model together with ComfyUI, a node based image generation tool. Rather than being an app someone runs directly, it is written to be loaded by an AI coding assistant, such as Claude Code or similar tools that can execute shell commands, which then follows the skill's instructions to carry out the full image generation process on the user's behalf. The workflow the skills guide an assistant through has four stages: planning the composition, meaning choosing framing, camera angle, and lighting based on a plain description, looking up valid tags through a separate command line tool, assembling a positive and negative prompt following the format Anima expects, and finally running the actual ComfyUI workflow to generate, upscale, and cache the image. The most important supporting piece is a tool called danbooru-tags, a command line program written in Rust that searches a local index built from Anima's official tag list. Because Anima was trained on the Danbooru tagging system, which has millions of possible tags, this tool checks whether an artist name, character name, or clothing tag the user asked for is actually valid, and suggests close matches when it is not, rather than letting the AI assistant guess and write an invalid tag that the model would simply ignore. Running this setup has real hardware requirements. It expects Windows 10 or 11, ComfyUI itself, Node.js, an NVIDIA GPU with at least 8 gigabytes of video memory though 12 or more is recommended, and a compatible version of CUDA and PyTorch. Python is only needed once, to build the initial tag search index, not for everyday image generation. A separate tool called comfyui-skill-cli acts as the bridge between the AI assistant and the locally running ComfyUI instance, and the README notes this bridge can also be used for other ComfyUI workflows beyond Anima. Several specific model files, including the Anima checkpoint, a text encoder, a VAE, and two LoRA files for higher resolution and quality, need to be downloaded separately and placed into ComfyUI's model folders before anything will generate. The README includes several sample images generated with this setup, credited to a specific artist style, and does not state a license.