ComfyUI Extension: Forbidden Vision

ComfyUI-Forbidden-Vision Introduction

ComfyUI-Forbidden-Vision is an innovative extension designed to enhance the capabilities of ComfyUI by providing advanced face detection and segmentation features. This extension is particularly useful for AI artists who work with both anime and realistic content, including both safe-for-work (SFW) and not-safe-for-work (NSFW) images. The primary goal of ComfyUI-Forbidden-Vision is to automatically detect faces in images and apply sophisticated corrections to improve the overall quality and appearance of the faces. This includes automatic face fixing, color and tone adjustments driven by learned models, and context-aware inpainting. By using this extension, AI artists can achieve more consistent and polished results in their digital artworks, saving time and effort in manual editing.

How ComfyUI-Forbidden-Vision Works

At its core, ComfyUI-Forbidden-Vision operates by leveraging custom-trained models to detect and segment faces within an image. Once a face is detected, the extension applies a series of automated processes to enhance the image. These processes include:

1. Detection & Segmentation: The extension uses specialized models to identify and segment faces, even in challenging scenarios such as extreme poses or heavy occlusion. This ensures that the detected faces are accurately isolated for further processing.
2. Neural Correction: After detection, the extension automatically analyzes the image to correct exposure, black levels, and color balance. This step is crucial for ensuring that the face appears natural and well-integrated with the rest of the image.
3. Inpainting and Refinement: The extension uses context-aware inpainting techniques to fill in any gaps or imperfections in the face, ensuring a seamless blend with the surrounding image. This is particularly useful for restoring details that may have been lost or distorted. By automating these processes, ComfyUI-Forbidden-Vision allows AI artists to focus on their creative work, while the extension handles the technical aspects of image enhancement.

ComfyUI-Forbidden-Vision Features

ComfyUI-Forbidden-Vision offers a range of features designed to enhance the quality of face images:

• Face Detection and Masking: Utilizes custom-trained YOLO11 and segmentation models to accurately detect and mask faces, ensuring precise inpainting results.
• NSFW Friendly: Capable of processing adult content without filtering, making it versatile for various artistic needs.
• Detail-Aware Segmentation: Recognizes facial details such as eyebrows, eyelashes, and sunglasses to improve inpainting accuracy.
• Context-Aware Inpainting: Matches the lighting and style of the original image for a cohesive final result.
• Flexible Prompt Control: Allows artists to add, replace, or exclude specific tags in prompts for face processing.
• AI Pre-Upscaling: Enhances small faces before processing to ensure detailed final results.
• Smart Blending: Applies automatic color correction and feathered mask blending for seamless integration.

ComfyUI-Forbidden-Vision Models

The extension includes three custom-trained models, each serving a specific purpose:

1. Detection & Segmentation Model: Ensures consistent face detection across various content types, including real, anime, and NSFW images.
2. Neural Corrector Model: Automatically adjusts exposure, black levels, and color balance to enhance the overall appearance of the face.
3. Inpainting Model: Fills in missing details and ensures a smooth transition between the face and the surrounding image. These models are designed to work together to provide a comprehensive solution for face enhancement in digital artworks.

Troubleshooting ComfyUI-Forbidden-Vision

While ComfyUI-Forbidden-Vision is designed to be user-friendly, you may encounter some common issues:

• Detection Failures: If the extension fails to detect a face, especially in highly stylized or occluded images, consider adjusting the detection confidence settings or manually rotating the image for better results.
• Color and Tone Issues: If the automatic corrections do not meet your expectations, you can adjust the neural corrector settings to fine-tune the tone and color balance.
• Inpainting Artifacts: In cases where inpainting results in visible artifacts, try adjusting the mask expansion and blending settings to achieve a more natural look. For further assistance, you can report issues on the GitHub Issues page.

Learn More about ComfyUI-Forbidden-Vision

To further explore the capabilities of ComfyUI-Forbidden-Vision, you can access additional resources such as tutorials, documentation, and community forums. These resources provide valuable insights and support for AI artists looking to maximize the potential of this extension. You can also explore the models used by the extension on HuggingFace, where they are available for download and experimentation.