Just a collection of modified, fixed, or newly created nodes for ComfyUI, aimed at enhancing the user experience and functionality of AI art workflows. This repository includes features like negative prompt weights and attention management, which improve the control and output quality of generated images.
- Supports advanced prompt weighting by allowing negative weights to negate specific traits in images.
- Offers a variety of custom nodes for managing attention, bounding boxes, and reference images, enhancing the flexibility of the ComfyUI environment.
- Provides compatibility with multiple models, ensuring that users can leverage the latest advancements in AI art generation.
Context
This tool serves as an extension for ComfyUI, providing a set of nodes that enhance the capabilities of AI art generation workflows. It primarily focuses on improving prompt control and attention management, enabling users to achieve more refined and tailored outputs.
Key Features & Benefits
The repository includes several unique nodes, such as CLIP NegPip, which allows for the use of negative weights in prompts to negate unwanted traits or concepts in generated images. Additionally, the AttentionCouplePPM node automates the management of new region inputs, streamlining the process of applying attention to specific areas of an image.
Advanced Functionalities
The tool features advanced capabilities like the Guidance Limiter, which applies guidance within a limited interval to enhance sample quality, and C2FG, which implements a time-dependent classifier-free guidance scale that improves image clarity and detail as the denoising process progresses. These functionalities allow for more nuanced control over the image generation process.
Practical Benefits
By integrating these nodes into their workflows, users can significantly improve the efficiency and quality of their image generation processes within ComfyUI. The ability to manage prompt weights and attention dynamically allows for more precise control over the final output, leading to higher-quality and more visually appealing results.
Credits/Acknowledgments
The original contributions come from various authors, including laksjdjf, hako-mikan, and ArcAnima. The tool is licensed under open-source terms, encouraging collaboration and further development within the community.




