ComfyUI Extension: Shibiko AI ComfyUI Tools

Shibiko AI ComfyUI Tools

What is this?

This is a collection of tools that I use to make my life easier when developing ComfyUI applications. Waifu2x to reduce noise and upscale 2x and 4x trained on art or phones. Luts used in the photo and film industry to recolor images. I use this tools to further develop features for Shibiko AI

Notes

I put in credits for the original creators of the underlining tools that I use to create nodes. So links in each section are for each creator, repo, and patreon if they have one.

Waifu2x

This tool uses the waifu2x AI model directly to upscale images and remove noise. It produces high quality upscaled images that are suitable for use in ComfyUI applications.

Credits: Nagadomi | GitHub | Pateron

Luts

I wrote this using built in methods from the PIL library and converting the image types. This comes straight from production code I use in Shibiko AI. If you have any issues, please check to make sure that you have a luts directory in the models directory with luts in it. A luts directory is auto created in the models directory where you can add your own luts.

Default Luts come from on1.com selected a few of the free ones to use in this tool. Please go to the site and check out the rest of the free luts.

Cascades

Wrote this to include the opencv cascades that are built in. I also include a cascade made by Nagadomi that you can find here: lbpcascade_animeface We can use this to detect faces, eyes, and anime faces without needing yolo. Yolo has an AGPL license, meaning you can't use it commerically unless you purchase a yearly license. Works well for characters and people that are facing towards the front. You can use this as an alternate way for inpainting or in combination with other masking techniques. A cascades directory is auto created in the models directory where you can add your own cascades.

Remove Noise

Advanced noise removal node using bilateral and guided filters from OpenCV. Provides fine-grained control over filter parameters including loop iterations, filter diameter, sigma values, and processing order. Perfect for cleaning up images while preserving edge details.

Credits: toyxyz | GitHub

LM Studio Vision

Vision language model (VLM) controller for processing images and video frames through locally-hosted LM Studio models. Features integrated ComfyUI progress tracking that shows which image is being processed during batch operations. Supports multiple image inputs, automatic frame sampling for videos, and flexible model configuration with GPU offloading and context length controls.

Key Features:

• Real-time Progress Tracking: Integrated ComfyUI progress bar shows current processing status (e.g., "Processing image 3/8")
• Batch Processing: Handles multiple images and video frames with automatic sampling
• Model Management: Automatic model loading/unloading with caching optimization
• Flexible Input: Supports up to 3 image inputs plus video frames
• Configurable Parameters: GPU offload ratio, context length, max tokens, temperature, and more

Requirements:

• LM Studio installed and running
• A vision-capable model loaded in LM Studio (e.g., LLaVA, Bakllava)

Credits: Original implementation by dandancow874 | GitHub | Enhanced with progress tracking for Shibiko AI Tools