Comfyui_TTP_Toolset

# ComfyUI Custom Nodes: Comfyui_TTP_Toolset

## Overview

The `Comfyui_TTP_Toolset` repository provides a collection of custom nodes specifically designed for enhancing image upscaling workflows in ComfyUI. By leveraging the DIT model and integrating various sampling and tile-based techniques, this toolset aims to deliver accurate and efficient image processing, minimizing hallucinations and maintaining image fidelity.

## Repository Purpose

The primary goal of this repository is to facilitate simple logic upscale nodes within the DIT model, supporting common applications such as Flux, Hunyuan, and SD3. The workflow divides the initial image into smaller tiles, extracts prompts using an image-interrogator for each tile, and subsequently performs a refined upscale process. This ensures precision in image processing and is particularly useful for handling large images.

## Node Offerings

This repository includes a variety of custom nodes, each serving specific functions within the image processing workflow:

- [**TTPlanet_Tile_Preprocessor_Simple**](/comfyui/custom-nodes/comfyui_ttp_toolset/ttplanet_tile_preprocessor_simple)
- [**TTP_Image_Tile_Batch**](/comfyui/custom-nodes/comfyui_ttp_toolset/ttp_image_tile_batch)
- [**TTP_Image_Assy**](/comfyui/custom-nodes/comfyui_ttp_toolset/ttp_image_assy)
- [**TTP_CoordinateSplitter**](/comfyui/custom-nodes/comfyui_ttp_toolset/ttp_coordinatesplitter)
- [**TTP_condtobatch**](/comfyui/custom-nodes/comfyui_ttp_toolset/ttp_condtobatch)
- [**TTP_condsetarea_merge**](/comfyui/custom-nodes/comfyui_ttp_toolset/ttp_condsetarea_merge)
- [**TTP_Tile_image_size**](/comfyui/custom-nodes/comfyui_ttp_toolset/ttp_tile_image_size)
- [**TTP_condsetarea_merge_test**](/comfyui/custom-nodes/comfyui_ttp_toolset/ttp_condsetarea_merge_test)
- [**TTP_Expand_And_Mask**](/comfyui/custom-nodes/comfyui_ttp_toolset/ttp_expand_and_mask)
- [**TTP_text_mix**](/comfyui/custom-nodes/comfyui_ttp_toolset/ttp_text_mix)
- [**TeaCacheHunyuanVideoSampler**](/comfyui/custom-nodes/comfyui_ttp_toolset/teacachehunyuanvideosampler)

## Special Features and Capabilities

### TeaCache Sampler Integration

The toolkit integrates the **TeaCache sampler** for the Hunyuan video, allowing for significant acceleration in video processing tasks. Leveraging code from the TeaCache repository, the sampler can dramatically reduce rendering times, though it may impact image quality and dynamics slightly.

- **Performance:** On an NVIDIA 4090, rendering a 720×480 resolution video with 65 frames completes in roughly 55 seconds, achieving a speedup factor of `x2.1`.

### Precision Support

The workflow supports `bf16` and `fp8` precision, ensuring compatibility with various performance demands.

### ControlNet Tile Integration

The workflow supports the **ControlNet Tile** feature, enhancing the upscale process when employing the Hunyuan DIT model. This integration is particularly useful for maintaining structural consistency across tiled sections.

## Usefulness in ComfyUI Workflows

This repository offers substantial enhancements for users seeking refined image upscaling within ComfyUI workflows. By integrating tile-based processing and advanced sampling techniques, the toolset ensures high-resolution results with minimized distortions and artifacts. It is especially beneficial for users handling complex and large-scale image processing projects.

## Installation

Currently, specific installation instructions are not provided in the repository. However, custom nodes are typically installed by cloning the repository into your ComfyUI project directory or adding it as a submodule.

For updates and additional details, please refer to the [GitHub repository](https://github.com/TTPlanetPig/Comfyui_TTP_Toolset).

--- 

Explore the comprehensive documentation of individual nodes linked within this document for a detailed understanding of their respective functionalities.