nodes/WJLUOXIAO-XB_ToolBox

Fork 0

Files

T

wjluoxiao 46e341fc00 Update README.md

2026-04-14 14:25:24 +08:00

8.4 KiB

Raw Blame History

README.md Markdown

🧰 ComfyUI XB-BOX (小白工具箱)

The Ultimate Spatiotemporal & VRAM Management Toolset for ComfyUI.

专为突破显存极限与 DiT 视频大模型（WAN 2.2 / LTX 2.3）深度定制的终极架构级扩展套件。

🇬🇧 English Documentation

🌟 Core Philosophy

The primary goal of XB_ToolBox is to help AI beginners quickly master ComfyUI workflows, making local deployment and operation simpler and more intuitive. It is a comprehensive ComfyUI extension suite rebuilt from the ground up, covering everything from low-level memory scheduling to front-end interactive UX.

It integrates commonly used but scattered parameter nodes, allowing users to set frequently used image and video parameters through a single, unified node.

It provides intuitive visual chunking and parameter preview nodes, helping beginners quickly understand the mechanics of encoder and decoder tiling/chunking.

It introduces a "Mirror Clone Dashboard" that simplifies highly complex workflows by cloning key operation widgets into a clean, centralized control panel!

Furthermore, through "Spatiotemporal Chunk Hijacking", "Dynamic VRAM Offloading for UNet/Checkpoints", and "Deep LiteGraph UI Customization", it enables consumer-grade GPUs with 24GB VRAM (especially in AMD 7900XTX ROCm environments) to smoothly run massive 14B~22B 3D-DiT video models that would normally cause instant Out-Of-Memory (OOM) crashes.

✨ Node Matrix

1. 🧊 Extreme VRAM & Model Optimizations

🧊 Sampler Chunk Master: Developed specifically for WAN/LTX models. Executes dual-axis slicing (Spatial Tiles & Temporal Chunks) directly in the latent space. It hijacks model parameters in place and features a built-in rocm_optimized extreme VRAM recovery strategy, completely breaking the curse of large-tensor OOM.
✂️ Model Block Swap: Supports both UNet and Checkpoint modes. Forcefully offloads the first N core blocks (Transformer Blocks) and Text/Image embedding layers to system RAM, trading generation time for VRAM space.
🧹 VRAM Cleaner: Essential for pre-generation clearing. Provides "nuclear-level cleanup" by deeply invoking gc.collect() and torch.cuda.empty_cache() to forcibly shatter PyTorch memory fragments.

2. 🎬 Media Params & Spatiotemporal Visualization

🎬 Media Params Master: Features three built-in modes: Free, Image, and Video. In video mode, it provides a geek-level "manual/automatic shifting engine" to smoothly switch between official pre-trained "golden bucket resolutions" (e.g., 480x832, 544x960, 720x1280), strictly locking to the 1+8N physically safe frame counts.
🧊 Chunk Visualization: Original dual-zone radar! The left side displays a 2D absolute hollow grid (S1, S2...), while the right side renders a temporal cylinder stack (T1, T2...) using a 3D painter's algorithm, intuitively visualizing the overlapping areas of all chunks.
📟 VRAM Calculator & Data Radar: Predicts VRAM occupation for WAN/LTX under different quantizations (FP8/GGUF) and provides exact tensor volume weighing down to the megabyte.

3. 🎛️ Ultimate Workflow UX/UI Enhancements

🪄 XB Dashboard Zen: A "Mirror Clone Dashboard" developed utilizing LiteGraph's underlying features. It allows you to batch "clone" any scattered widgets in the workflow into a fixed, central panel. Featuring two-way data synchronization and direct pipeline connections to build your personalized control center.
🎛️ Dynamic Bus: An extremely compact N-in, N-out AnyType universal node. Features UI-customizable type labels and one-click channel addition/removal, saving you from messy, tangled noodle wires.

📦 Installation

Method 1: Via ComfyUI ManagerSearch for XB_ToolBox in the ComfyUI Manager and click install.

Method 2: Manual Git Installation
Navigate to your ComfyUI custom_nodes directory and run the following command:

git clone https://github.com/wjluoxiao/XB_ToolBox.git

(Note: This plugin relies purely on the native ComfyUI ecosystem. Implemented through an elegant Python/JS architecture, no additional pip dependencies are required! Plug and play.)

📝 License

This project is open-sourced under the Apache-2.0 License. While embracing extreme freedom of sharing, it ensures patent defense rights for the core architecture code.

🇨🇳 中文说明 (Chinese)

🌟 核心理念

XB_ToolBox 的主旨是帮助新手AI玩家快速上手ComfyUI工作流，让在本地部署运行更简单方便，是一套从底层内存调度到前端交互体验进行全方位重构的 ComfyUI 插件。

它整合了工作流常用但分散的参数节点，让大家可以通过一个节点设置常用的图片视频参数。

提供了更直观的可视化分块预览和参数预览节点，让新手更快理解解码器和编码器分块的原理。

提供了“镜像克隆台”这种可以把超级复杂的工作流，通过克隆主要操作节点，从而让工作流更加简洁的方案！

另一方面也通过 “时空分块劫持”、“UNet / Checkpoint 动态显存外包” 以及 “LiteGraph 交互深度定制”，让拥有 24G 显存（尤其是 AMD 7900XTX 等 ROCm 环境）的消费级显卡，也能流畅运行原本会直接 OOM 的 14B~22B 超大 3D-DiT 视频模型。

✨ 核心节点矩阵 (Node Matrix)

1. 🧊 极限显存特攻 (VRAM & Model Optimizations)

🧊 采样分块大师 (Sampler Chunk Master)：专为 WAN/LTX 开发。在潜空间（Latent）执行时空双轴切割（Spatial Tiles & Temporal Chunks）。原地劫持模型参数，内置 rocm_optimized 极致显存回收策略，彻底打破大张量 OOM 魔咒。
✂️ 模型分块交换 (Block Swap)：支持 UNet 与 Checkpoint 模式。可将前 N 个核心块（Transformer Blocks）以及 Text/Image 嵌入层强制卸载至系统内存（RAM），用时间换取空间。
🧹 显存清理大师 (VRAM Cleaner)：战前清场必备。提供“核爆级清理”，深度调用 gc.collect() 与 torch.cuda.empty_cache()，强行粉碎 PyTorch 缓存碎片。

2. 🎬 媒体参数与时空可视化 (Media Params & Visualization)

🎬 图像参数大全 (Media Params Master)：内置“自由/图片/视频”三大模式。在视频模式下，提供极客级“手自一体换挡引擎”，在官方预训练的“黄金桶分辨率”（如 480x832, 544x960, 720x1280）之间平滑切挡，并严格锁定 1+8N 物理安全帧数。
🧊 时空分块预览 (Chunk Visualization)：独创双区雷达！左侧展示 2D 绝对镂空网格（S1, S2...），右侧通过 3D 画家算法渲染时间圆柱堆叠（T1, T2...），直观展示所有块的重叠区。
📟 可用显存计算 & 数据雷达：预判 WAN/LTX 在不同量化（FP8/GGUF）下的显存驻留，并提供精确到 MB 级别的张量体积称重。

3. 🎛️ 终极工作流 UI (Workflow UX/UI Enhancements)

🪄 XB 远程控制中心 (Dashboard Zen)：利用 LiteGraph 底层特性开发的“镜像克隆台”。可将工作流中分散的任意组件批量“克隆”到一个固定面板中。双向数据同步，大动脉直连，打造属于你的定制化监控台。
🎛️ 动态总线 (Dynamic Bus)：极度压缩尺寸的 N进N出 AnyType 万能节点。提供 UI 自定义类型标签与一键通道增减，拯救杂乱无章的节点面条线。

📦 安装指南 (Installation)

方法一：通过 ComfyUI Manager

搜索 XB_ToolBox 或 小白工具箱 进行安装。

方法二：手动 Git 安装

进入你的 ComfyUI custom_nodes 目录，执行以下命令：

Bash

git clone https://github.com/wjluoxiao/XB_ToolBox.git

(注：本插件纯粹依赖 ComfyUI 原生生态，通过精妙的 Python/JS 架构实现，无需安装任何额外 pip 依赖！ 即插即用。)

📝 许可证 (License)

本项目基于 Apache-2.0 License 开源。在极度自由分享的同时，保障了核心架构代码的专利防御权。

8.4 KiB Raw Blame History