Baidu Meux ComfyTools

English | 中文

---

English

Overview

Baidu Meux ComfyTools is a collection of custom nodes for ComfyUI used by the Baidu Meux asset platform.
Current version: 1.7.1

Main nodes:

• MeuxImageLoader: load an image from ComfyUI/input or an HTTP/HTTPS URL.
• MeuxMaskLoader: load a mask from ComfyUI/input or an HTTP/HTTPS URL and align it to a reference image.
• MeuxMaskFillHoles: fill enclosed holes in a mask.
• MeuxMaskBlurPlus: blur or feather a mask.
• MeuxSeed: workflow-compatible seed passthrough node.
• MeuxMultiSaveImage: save up to sixteen image batches with optional resizing.
• MeuxAdvancedImageCrop: crop images by pixels or percentage with optional alignment.
• MeuxSimpleLLMNode: call a chat-completions style LLM API inside a workflow.
• MeuxTextAreaInput: multi-line text input helper.
• MeuxArtisticTextPreview: advanced transparent text preview with fill, outline, spacing, shadow, glow, and mask output.
• MeuxSmartEmptyLatent: generate a safe-sized empty latent tensor.
• MeuxSizePresetSafe: compute aligned generation sizes.
• MeuxOutpaintSizePresetSafe: compute safe per-side outpaint expansion sizes.
• MeuxSmartExactResize: crop or pad to an exact target size.
• MeuxRMBG: BiRefNet-based background removal.
• MeuxRealESRGANUpscale: local RealESRGAN upscaling.

Workflow-compatible aliases:

• Mask Fill Holes
• MaskBlur+
• Seed

Changelog

• v1.7.1
  • Updated URL persistence so downloaded files are stored under ComfyUI/input/meux_downloads.
  • Added text_align and vertical_align controls to MeuxArtisticTextPreview.
• v1.7.0
  • Added MeuxArtisticTextPreview for transparent text preview rendering with fill, outline, spacing, mask output, and preset-based effects.
  • Added a project-level fonts/ directory and made it the first font lookup location for the preview node.
  • Added node-specific parameter documentation for MeuxArtisticTextPreview.
• v1.6.0
  • Added MeuxMaskLoader as a dedicated mask-loading node.
  • Added workflow-compatible mask utility nodes: MeuxMaskFillHoles, MeuxMaskBlurPlus, MeuxSeed.
  • Registered compatibility aliases: Mask Fill Holes, MaskBlur+, Seed.
  • Added scipy dependency for hole-filling behavior aligned with upstream workflows.
• v1.5.0
  • Added MeuxRMBG for local/remote BiRefNet background removal.
  • Added MeuxRealESRGANUpscale for local RealESRGAN upscaling.
• v1.4.0
  • Added MeuxOutpaintSizePresetSafe.
  • Capped size-related inputs at 4096.
• v1.3.0
  • Added MeuxSmartEmptyLatent, MeuxSizePresetSafe, and MeuxSmartExactResize.
• v1.2.0
  • Added MeuxImageLoader with URL downloading support.
• v1.1.0
  • Restructured the package into modular files under nodes/.
• v1.0.0
  • Initial release with MeuxMultiSaveImage.

Installation / Update

1. Clone into ComfyUI/custom_nodes/:

   cd ComfyUI/custom_nodes
   git clone https://github.com/fchangjun/Baidu_Meux_ComfyTools.git

2. Update when needed:

   cd ComfyUI/custom_nodes/Baidu_Meux_ComfyTools
   git pull

3. Install dependencies:

   cd ComfyUI/custom_nodes/Baidu_Meux_ComfyTools
   pip install -r requirements.txt

4. Restart ComfyUI.

Quick Usage

MeuxImageLoader

• source_type = local: select a file from ComfyUI/input
• source_type = url: download from an HTTP/HTTPS URL
• Downloaded URL files are persisted under ComfyUI/input/meux_downloads when persist_to_input = true
• Outputs IMAGE and the derived mask from alpha or grayscale when present

MeuxMaskLoader

• reference_image is required so the mask can match the image size
• source_type = local | url
• Supports mask_channel, mask_invert, and resize_to_reference
• Outputs native ComfyUI-style MASK

Recommended pairing:

1. MeuxImageLoader.image -> downstream image pipeline
2. MeuxImageLoader.image -> MeuxMaskLoader.reference_image
3. MeuxMaskLoader.mask -> downstream mask pipeline

Workflow-Compatible Mask Nodes

• MeuxMaskFillHoles / Mask Fill Holes: fill enclosed regions inside a mask
• MeuxMaskBlurPlus / MaskBlur+: feather a mask
• MeuxSeed / Seed: expose a workflow-compatible seed value

Other Utility Nodes

• MeuxMultiSaveImage: batch image saving
• MeuxAdvancedImageCrop: crop by pixels or percentage
• MeuxSimpleLLMNode: direct LLM API calls
• MeuxTextAreaInput: long prompt input
• MeuxArtisticTextPreview: advanced transparent text preview with optional gradient fill, outline, shadow, inner shadow, glow, and mask output
• MeuxSmartEmptyLatent: safe latent initialization
• MeuxSizePresetSafe: safe size planning
• MeuxOutpaintSizePresetSafe: safe outpaint size planning
• MeuxSmartExactResize: exact resize with crop/pad logic
• MeuxRMBG: local background removal
• MeuxRealESRGANUpscale: local image upscaling

MeuxArtisticTextPreview

• Outputs a transparent IMAGE preview and a matching MASK
• Reads fonts from Baidu_Meux_ComfyTools/fonts first, then falls back to bundled/system fonts
• Supports fixed and fit font sizing modes
• Supports font_name, char_spacing, bold, and italic
• Supports solid or gradient fill
• Supports solid or gradient outline
• Supports shadow, inner shadow, and outer glow
• Supports transparent or colored preview backgrounds

Recommended first test:

1. Add MeuxArtisticTextPreview
2. Set text = Hello
3. Set font_size_mode = fixed
4. Set font_size = 128
5. Set outline_type = solid
6. Connect the image output to a preview or save node

Parameters:

• Basic

  • text: preview text, supports multiple lines
  • font_name: font dropdown, reads project fonts/ first
  • width / height: output canvas size
  • font_size_mode: fixed uses the exact font size, fit scales text to the canvas
  • font_size: base font size or max fitting size
  • char_spacing: character spacing in pixels
  • line_spacing: line spacing in pixels; 0 keeps the automatic spacing rule, positive values increase spacing, negative values pull lines closer together
  • bold: prefer bold font variants, otherwise use a fallback bold draw mode
  • italic: prefer italic font variants, otherwise use a fallback italic transform
  • background_color: preview background color
  • background_opacity: preview background opacity, 0 means fully transparent
  • padding_percent: safe area margin inside the canvas

• Fill

  • fill_type: solid, gradient, or none
  • fill_color_1: solid fill color, or the start color of a gradient
  • fill_color_2: end color of a gradient
  • fill_direction: fill gradient direction

• Outline

  • outline_type: none, solid, or gradient
  • outline_width: outline thickness
  • outline_opacity: outline alpha
  • outline_color_1: solid outline color, or the start color of an outline gradient
  • outline_color_2: end color of an outline gradient
  • outline_direction: outline gradient direction

• Effects

  • effect_preset: one-click effect preset
  • effect_strength: preset intensity multiplier
  • shadow_mode: preset uses preset shadow settings, custom ignores preset shadow placement/blur and uses manual values below
  • shadow_color: shadow color; still works in preset mode
  • shadow_offset_x: custom shadow horizontal offset
  • shadow_offset_y: custom shadow vertical offset
  • shadow_opacity: custom shadow alpha
  • shadow_blur: custom shadow blur radius; 0 means a solid hard-edged shadow, > 0 means a blurred shadow
  • Presets:
    • none: no extra effect
    • soft_shadow: light shadow
    • heavy_shadow: heavier shadow
    • soft_glow: soft glow
    • neon_glow: stronger glow
    • engraved: inner-shadow style engraved look
    • shadow_glow: combines shadow and glow

• Outputs

  • image: rendered preview image
  • mask: alpha mask derived from the text result

Folder Structure

Baidu_Meux_ComfyTools/
├── __init__.py
├── fonts/
│   └── README.md
├── pyproject.toml
├── requirements.txt
└── nodes/
    ├── advanced_image_crop.py
    ├── image_loader.py
    ├── mask_blur_plus.py
    ├── mask_fill_holes.py
    ├── mask_loader.py
    ├── artistic_text_preview.py
    ├── multi_save_image.py
    ├── outpaint_size_preset_safe.py
    ├── realesrgan_upscale.py
    ├── rmbg_birefnet.py
    ├── seed_node.py
    ├── simple_llm_node.py
    ├── size_preset_safe.py
    ├── smart_empty_latent.py
    ├── smart_exact_resize.py
    └── text_area_input.py

Requirements

• ComfyUI
• PyTorch
• Pillow
• NumPy
• Requests
• transformers
• torchvision
• realesrgan
• scipy

Support

• Repository: fchangjun/Baidu_Meux_ComfyTools
• Issues: GitHub Issues

---

中文

概述

Baidu Meux ComfyTools 是一组用于 ComfyUI 的自定义节点，服务于百度 Meux 资产平台相关工作流。
当前版本：1.7.1

主要节点：

• MeuxImageLoader：从 ComfyUI/input 或 HTTP/HTTPS URL 加载图片
• MeuxMaskLoader：从 ComfyUI/input 或 HTTP/HTTPS URL 加载遮罩，并对齐到参考图尺寸
• MeuxMaskFillHoles：填充遮罩内部封闭空洞
• MeuxMaskBlurPlus：对遮罩做模糊/羽化
• MeuxSeed：兼容工作流的种子透传节点
• MeuxMultiSaveImage：一次保存最多 16 组图像，可选统一尺寸
• MeuxAdvancedImageCrop：按像素或百分比裁剪图像
• MeuxSimpleLLMNode：在工作流中调用聊天式 LLM 接口
• MeuxTextAreaInput：多行文本输入辅助节点
• MeuxArtisticTextPreview：在透明画布上渲染高级文字预览，支持填充、描边、字距、阴影、内阴影、外发光，并输出遮罩
• MeuxSmartEmptyLatent：生成安全尺寸的空白 latent
• MeuxSizePresetSafe：计算安全生成尺寸
• MeuxOutpaintSizePresetSafe：计算安全外扩尺寸
• MeuxSmartExactResize：精确尺寸裁剪/补边
• MeuxRMBG：基于 BiRefNet 的抠图节点
• MeuxRealESRGANUpscale：本地 RealESRGAN 放大节点

工作流兼容别名：

• Mask Fill Holes
• MaskBlur+
• Seed

更新日志

• v1.7.1
  • 调整 URL 落盘位置，下载文件现在保存到 ComfyUI/input/meux_downloads
  • 为 MeuxArtisticTextPreview 新增 text_align 和 vertical_align 对齐控制
• v1.7.0
  • 新增 MeuxArtisticTextPreview，用于透明背景文字预览，支持填充、描边、字距、遮罩输出和预设效果。
  • 新增项目级 fonts/ 目录，并作为预览节点的第一字体读取位置。
  • 补充 MeuxArtisticTextPreview 的节点参数说明文档。
• v1.6.0
  • 新增独立遮罩加载节点 MeuxMaskLoader
  • 新增工作流兼容节点 MeuxMaskFillHoles、MeuxMaskBlurPlus、MeuxSeed
  • 注册兼容别名 Mask Fill Holes、MaskBlur+、Seed
  • 新增 scipy 依赖，用于按上游逻辑填充遮罩空洞
• v1.5.0
  • 新增 MeuxRMBG，支持本地/远端 BiRefNet 抠图
  • 新增 MeuxRealESRGANUpscale，支持本地 RealESRGAN 放大
• v1.4.0
  • 新增 MeuxOutpaintSizePresetSafe
  • 尺寸相关输入上限统一为 4096
• v1.3.0
  • 新增 MeuxSmartEmptyLatent、MeuxSizePresetSafe、MeuxSmartExactResize
• v1.2.0
  • 新增 MeuxImageLoader，支持 URL 下载
• v1.1.0
  • 重构为 nodes/ 模块化结构
• v1.0.0
  • 发布 MeuxMultiSaveImage 初版

安装 / 更新

1. 克隆到 ComfyUI/custom_nodes/：

   cd ComfyUI/custom_nodes
   git clone https://github.com/fchangjun/Baidu_Meux_ComfyTools.git

2. 更新代码：

   cd ComfyUI/custom_nodes/Baidu_Meux_ComfyTools
   git pull

3. 安装依赖：

   cd ComfyUI/custom_nodes/Baidu_Meux_ComfyTools
   pip install -r requirements.txt

4. 重启 ComfyUI。

快速使用

MeuxImageLoader

• source_type = local：从 ComfyUI/input 选择本地文件
• source_type = url：运行时下载 HTTP/HTTPS 图片
• persist_to_input = true 时，URL 下载文件会保存到 ComfyUI/input/meux_downloads
• 输出 IMAGE，以及图片自带 alpha/灰度派生的 MASK

MeuxMaskLoader

• 必须连接 reference_image，用于保证遮罩尺寸和原图一致
• source_type = local | url
• 支持 mask_channel、mask_invert、resize_to_reference
• 输出原生兼容的 MASK

推荐连法：

1. MeuxImageLoader.image 接图像处理链路
2. MeuxImageLoader.image 接 MeuxMaskLoader.reference_image
3. MeuxMaskLoader.mask 接遮罩处理链路

工作流兼容遮罩节点

• MeuxMaskFillHoles / Mask Fill Holes：填充遮罩内部封闭区域
• MeuxMaskBlurPlus / MaskBlur+：做遮罩羽化
• MeuxSeed / Seed：输出兼容工作流的 seed

其他工具节点

• MeuxMultiSaveImage：批量保存图片
• MeuxAdvancedImageCrop：按像素或百分比裁剪
• MeuxSimpleLLMNode：直接调用 LLM API
• MeuxTextAreaInput：长文本输入
• MeuxArtisticTextPreview：高级透明文字预览节点，支持渐变填充、描边、字距、阴影、内阴影、外发光和遮罩输出
• MeuxSmartEmptyLatent：安全 latent 初始化
• MeuxSizePresetSafe：安全尺寸规划
• MeuxOutpaintSizePresetSafe：安全外扩尺寸规划
• MeuxSmartExactResize：精确尺寸裁剪/补边
• MeuxRMBG：本地抠图
• MeuxRealESRGANUpscale：本地放大

MeuxArtisticTextPreview

• 输出透明背景 IMAGE 预览图，以及对应的 MASK
• 优先读取 Baidu_Meux_ComfyTools/fonts 目录中的字体，再回退到内置字体和系统字体
• 支持 fixed 和 fit 两种字号模式
• 支持 font_name、char_spacing、bold、italic
• 支持纯色或渐变填充
• 支持纯色或渐变描边
• 支持阴影、内阴影、外发光
• 支持透明或纯色背景预览

建议先用下面这组参数测试：

1. 添加 MeuxArtisticTextPreview
2. 设置 text = Hello
3. 设置 font_size_mode = fixed
4. 设置 font_size = 128
5. 设置 outline_type = solid
6. 将 image 输出连接到预览或保存节点

参数说明：

• 基础参数

  • text：预览文字，支持多行
  • font_name：字体下拉，优先读取项目根目录 fonts/
  • width / height：输出画布尺寸
  • font_size_mode：fixed 使用固定字号，fit 自动适配画布
  • font_size：基础字号，或自动适配时的最大字号
  • char_spacing：字符间距，单位为像素
  • line_spacing：行间距，单位为像素；0 表示继续使用自动行间距规则，正数拉开行距，负数压缩行距
  • bold：优先匹配粗体字体，没有时走伪粗体回退
  • italic：优先匹配斜体字体，没有时走伪斜体回退
  • background_color：预览背景颜色
  • background_opacity：背景透明度，0 表示完全透明
  • padding_percent：文字与画布边缘的安全边距比例

• 填充参数

  • fill_type：solid 纯色，gradient 渐变，none 无填充
  • fill_color_1：纯色填充时使用它；渐变时是起始色
  • fill_color_2：渐变结束色
  • fill_direction：填充渐变方向

• 描边参数

  • outline_type：none 无描边，solid 纯色描边，gradient 渐变描边
  • outline_width：描边宽度
  • outline_opacity：描边透明度
  • outline_color_1：纯色描边时使用它；渐变时是起始色
  • outline_color_2：渐变结束色
  • outline_direction：描边渐变方向

• 效果参数

  • effect_preset：一键效果预设
  • effect_strength：预设效果强度倍率
  • shadow_mode：preset 使用预设阴影参数，custom 使用下面的自定义阴影参数
  • shadow_color：阴影颜色；在预设模式下也会生效
  • shadow_offset_x：自定义阴影水平偏移
  • shadow_offset_y：自定义阴影垂直偏移
  • shadow_opacity：自定义阴影透明度
  • shadow_blur：自定义阴影模糊半径；0 表示实心硬边阴影，> 0 表示模糊阴影
  • 预设说明：
    • none：不加额外效果
    • soft_shadow：轻阴影
    • heavy_shadow：重阴影
    • soft_glow：柔和发光
    • neon_glow：更强的霓虹发光
    • engraved：偏内阴影的雕刻感
    • shadow_glow：阴影和发光组合

• 输出

  • image：渲染后的预览图
  • mask：由文字 alpha 生成的遮罩

目录结构

Baidu_Meux_ComfyTools/
├── __init__.py
├── fonts/
│   └── README.md
├── pyproject.toml
├── requirements.txt
└── nodes/
    ├── advanced_image_crop.py
    ├── image_loader.py
    ├── mask_blur_plus.py
    ├── mask_fill_holes.py
    ├── mask_loader.py
    ├── artistic_text_preview.py
    ├── multi_save_image.py
    ├── outpaint_size_preset_safe.py
    ├── realesrgan_upscale.py
    ├── rmbg_birefnet.py
    ├── seed_node.py
    ├── simple_llm_node.py
    ├── size_preset_safe.py
    ├── smart_empty_latent.py
    ├── smart_exact_resize.py
    └── text_area_input.py

依赖

• ComfyUI
• PyTorch
• Pillow
• NumPy
• Requests
• transformers
• torchvision
• realesrgan
• scipy

支持

• 仓库地址：fchangjun/Baidu_Meux_ComfyTools
• 问题反馈：GitHub Issues