This project provides a ComfyUI + Anima anime image generation skill package designed for AI programming assistants, offering precise control over image generation through a structured prompt system and robust tag validation. It aims to return creative control to the model and user by focusing on core generation constraints, enabling AI agents to generate high-quality anime art with greater accuracy and intent.

How It Works

The "v2mini" architecture employs a three-stage pipeline: comfyui-animatool for initial prompt and argument generation based on contextual causality, danbooru-tags for precise validation of Danbooru hard anchors (characters, artists, etc.) using a Rust CLI, and comfyui-manager for executing the ComfyUI workflow. This approach prioritizes "情境因果 → 三层 prompt → 精确执行" (contextual causality → three-layer prompt → precise execution), breaking prompts into hard_tags, soft_phrases, and nltags_block for granular control and avoiding invalid tags.

Quick Start & Requirements

• Installation: Clone ComfyUI, install comfyui-skill-cli (pip install comfyui-skill-cli), and clone necessary custom nodes into ComfyUI/custom_nodes/. Set the COMFYUI_GOOD_ANIMA_SKILLS_DIR environment variable to the root of the cloned comfyui-good-anima repository. Import workflows using comfyui-skill workflow import.
• Prerequisites: Windows 10/11, ComfyUI (latest), comfyui-skill-cli (latest), Node.js (18+), Python (3.10+ for index init), NVIDIA GPU (8GB+ VRAM, 12GB+ recommended), CUDA (12.8+), PyTorch (compatible with CUDA 12.8), xformers (0.0.3.0). Specific custom nodes are required (e.g., ANIMA_BOOSTER, FLSamplerV4, RES4LYF).
• Setup Time: Estimated setup involves cloning repositories, installing dependencies, and configuring environment variables, likely taking 30-60 minutes depending on user familiarity.
• Links:
  • ComfyUI: https://github.com/comfyanonymous/ComfyUI
  • ComfyUI Skill CLI: https://github.com/HuangYuChuh/ComfyUI_Skill_CLI
  • Custom Nodes: Links provided within the README for each required node.

Highlighted Details

• danbooru-tags Rust CLI: A core component for high-speed retrieval and exact/prefix validation of Danbooru tags, crucial for precise content control.
• Three-Layer Prompt System: Organizes prompts into hard_tags (confirmed anchors), soft_phrases (aesthetic terms), and nltags_block (spatial, causal, lighting details).
• AI Agent Compatibility: Designed for seamless integration with AI coding assistants like Snow, Claude Code, Codex, and PI, provided they can execute shell commands.
• Contextual Causality First: Prioritizes establishing the narrative and causal relationships within a scene before detailing visual attributes.

Maintenance & Community

The project focuses on maintaining the v2mini branch as the primary active development line, archiving older versions. While specific community channels like Discord or Slack are not listed, the project relies on and acknowledges contributions from various GitHub repositories and developers.

Licensing & Compatibility

The project is licensed under GPLv3. This copyleft license may impose restrictions on linking with closed-source software or commercial use without adhering to its terms.

Limitations & Caveats

Image-to-image generation is not integrated into the default pipeline. The "enhancer" workflow may introduce sharpening or quality degradation as a side effect of its speed optimization. The GPLv3 license requires careful consideration for commercial applications.