marhensa/comfyui-workflow

---
language:
- en
tags:
- comfyui
- template
- workflow
---
# HOW TO USE:

Drag and Drop Template.JSON / Template.PNG into [ComfyUI in browser](https://github.com/comfyanonymous/ComfyUI) to load the template! (Yes even output PNG file works as workflow template).

# PREQUISITE INSTALLATION NOTES:

It requires custom nodes and file requirements, it's easy:

1. ["ComfyUI Manager"](https://github.com/ltdrdata/ComfyUI-Manager#installation) (install it first).
2. Open ComfyUI WebUI > Manager > Install Custom Nodes, search "ComfyUI-Impact-Pack", "SDXL Prompt Sytler", and "Recommended SDXL Resolution Calculator", "Masquerade Nodes", Install ALL, close the ComfyUI console command prompt.
3. "Recommended SD 1.5 Resolution Calculator", open command prompt from this folder \ComfyUI\custom_nodes\ then use this command "git clone https://github.com/marhensa/sd15-recommended-res-calc.git")
4. "ComfyUI ControlNet Aux" custom node can be installed with [this instruction](https://github.com/Fannovel16/comfyui_controlnet_aux#installation), and the models can be downloaded [from here](https://huggingface.co/lllyasviel/sd_control_collection/tree/main), put into \ComfyUI\models\controlnet\
5. LCM LoRA, (needs a LCM LoRA, download it from here for [LCM-SDXL](https://huggingface.co/latent-consistency/lcm-lora-sdxl/resolve/main/pytorch_lora_weights.safetensors?download=true), for [LCM-SDXL-SSD1B](https://huggingface.co/latent-consistency/lcm-lora-ssd-1b/resolve/main/pytorch_lora_weights.safetensors?download=true), and here for [LCM-SD1.5](https://huggingface.co/latent-consistency/lcm-lora-sdv1-5/resolve/main/pytorch_lora_weights.safetensors?download=true)), for LCM-SDXL put into \ComfyUI\models\loras\LCM\SDXL\, LCM-SDXL-SSD1B put into \ComfyUI\models\loras\LCM\SDXL-SSD1B\, LCM-SD1.5 put into \ComfyUI\models\loras\LCM\SD15\ .
6. LCM-Turbo, download from [here](https://civitai.com/api/download/models/246747?type=Model&format=SafeTensor) for SDXL based models, and then put into put into \ComfyUI\models\loras\SDXL\LCMTurboMix_Euler_A_fix.safetensors
7. Another LoRA, download additionals LoRA from CivitAI or select "None" to not use it.
8. VAE selector, (needs a VAE file, download SDXL BF16 VAE from [here](https://huggingface.co/madebyollin/sdxl-vae-fp16-fix/resolve/main/sdxl_vae.safetensors), and VAE file for SD 1.5 from [here](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/resolve/main/vae-ft-mse-840000-ema-pruned.safetensors?download=true). Put into \ComfyUI\models\vae\SDXL\ and \ComfyUI\models\vae\SD15\).
9. Upscale model, (needs to be downloaded into \ComfyUI\models\upscale_models\ Recommended one is 4x-UltraSharp, download [from here](https://mega.nz/folder/qZRBmaIY#nIG8KyWFcGNTuMX_XNbJ_g). All the list of Upscale model is [here](https://upscale.wiki/w/index.php?title=Model_Database&oldid=1571))
10. Checkpoints. SDXL-SSD1B can be downloaded from [here](https://huggingface.co/segmind/SSD-1B/resolve/main/SSD-1B.safetensors?download=true), my recommended Checkpoint for SDXL is [AlbedoBase-XL](https://civitai.com/models/140737), [Jib-Mix-Realistic-XL](https://civitai.com/models/194768), [Crystal-Clear-XL](https://civitai.com/models/122822), and for SD1.5 is [Haveall](https://civitai.com/models/213692), download Safetensors file and put into \ComfyUI\models\checkpoints\SDXL\ and \ComfyUI\models\checkpoints\SD15\
11. Note: When loading a PNG Workflow from here, first click Refresh on ComfyUI menu, it will refresh models that you have on your PC, then choose it (Checkpoints, VAE, LoRA, Upscale Model, ControlNet models etc) accordingly.

.

> Custom Node "[Recommended Resolution Calculator](https://github.com/marhensa/sdxl-recommended-res-calc)", makes targeting final resolution much easier, because it will calculates recommended SD initial size and its upscale (or reverse upscale) value.

You only need to input your FINAL target resolution with this custom node.

.

# WORKFLOW SELECTION:
(drag and drop that PNG image file into ComfyUI interface, it will open up as workflow template)

## RECOMMENDED! FAST!! TIDY - Single SDXL Checkpoint Workflow (LCM-Turbo, PromptStyler, Upscale Model Switch, ControlNet, FaceDetailer) :

(ControlNet image reference example: [halo.jpg](https://huggingface.co/datasets/marhensa/comfyui-workflow/resolve/main/halo.jpg))

.

## TIDY - Single SD 1.5 Checkpoint Workflow (LCM, PromptStyler, Upscale Model Switch, ControlNet, FaceDetailer) :

(ControlNet image reference example: [halo.jpg](https://huggingface.co/datasets/marhensa/comfyui-workflow/resolve/main/halo.jpg))

.

## Breakdown - SDXL Workflow (LCM-Turbo, PromptStyler, Upscale Model Switch) :

.

## Breakdown - SDXL-SSD1B Workflow (LCM, PromptStyler, Upscale Model Switch) :

.

## Breakdown - SD1.5 Workflow (LCM, PromptStyler, Upscale Model Switch) :

.

## Simple without install anything - SDXL / SD15 LCM Workflow :

language:
- 英语
tags:
- ComfyUI
- 模板
- 工作流

# 使用方法：

将Template.JSON或Template.PNG拖拽至[浏览器版ComfyUI](https://github.com/comfyanonymous/ComfyUI)即可加载该模板！（即便输出的PNG文件也可作为工作流模板使用）

# 前置安装说明：

本数据集依赖自定义节点与文件资源，安装流程简便如下：

1. **ComfyUI管理器（ComfyUI Manager）**：需先行安装，安装指南见[此处](https://github.com/ltdrdata/ComfyUI-Manager#installation)。
2. 打开ComfyUI网页端（WebUI），进入「管理器」（Manager）板块，选择「安装自定义节点」，搜索「ComfyUI-Impact-Pack」「SDXL提示词风格器（SDXL Prompt Styler）」「推荐SDXL分辨率计算器」「遮罩节点（Masquerade Nodes）」并全部安装，完成后关闭ComfyUI的控制台命令窗口。
3. 「推荐SD 1.5分辨率计算器」：前往`ComfyUIcustom_nodes`文件夹，打开命令提示符，执行命令`git clone https://github.com/marhensa/sd15-recommended-res-calc.git`完成安装。
4. ComfyUI ControlNet Aux自定义节点：可按照[此指南](https://github.com/Fannovel16/comfyui_controlnet_aux#installation)完成安装，其依赖的模型可从[此处](https://huggingface.co/lllyasviel/sd_control_collection/tree/main)下载，并放置至`ComfyUImodelscontrolnet`文件夹中。
5. LCM LoRA（低秩适配器）：需下载对应版本的LCM LoRA：
- LCM-SDXL：[下载链接](https://huggingface.co/latent-consistency/lcm-lora-sdxl/resolve/main/pytorch_lora_weights.safetensors?download=true)，放置至`ComfyUImodelslorasLCMSDXL`
- LCM-SDXL-SSD1B：[下载链接](https://huggingface.co/latent-consistency/lcm-lora-ssd-1b/resolve/main/pytorch_lora_weights.safetensors?download=true)，放置至`ComfyUImodelslorasLCMSDXL-SSD1B`
- LCM-SD1.5：[下载链接](https://huggingface.co/latent-consistency/lcm-lora-sdv1-5/resolve/main/pytorch_lora_weights.safetensors?download=true)，放置至`ComfyUImodelslorasLCMSD15`
6. LCM-Turbo：适用于SDXL系列模型的版本可从[此处](https://civitai.com/api/download/models/246747?type=Model&format=SafeTensor)下载，并放置至`ComfyUImodelslorasSDXLLCMTurboMix_Euler_A_fix.safetensors`
7. 其他LoRA：可从CivitAI平台下载额外LoRA，或选择「无」以不使用额外LoRA。
8. VAE选择器：需下载对应VAE文件：
- SDXL BF16 VAE：[下载链接](https://huggingface.co/madebyollin/sdxl-vae-fp16-fix/resolve/main/sdxl_vae.safetensors)，放置至`ComfyUImodelsvaeSDXL`
- SD 1.5 VAE：[下载链接](https://huggingface.co/stabilityai/sd-vae-ft-mse-original/resolve/main/vae-ft-mse-840000-ema-pruned.safetensors?download=true)，放置至`ComfyUImodelsvaeSD15`
9. 超分模型：需下载至`ComfyUImodelsupscale_models`文件夹，推荐使用4x-UltraSharp模型，[下载链接](https://mega.nz/folder/qZRBmaIY#nIG8KyWFcGNTuMX_XNbJ_g)，完整超分模型列表可查阅[此处](https://upscale.wiki/w/index.php?title=Model_Database&oldid=1571)
10. 检查点（Checkpoint）模型：
- SDXL-SSD1B：[下载链接](https://huggingface.co/segmind/SSD-1B/resolve/main/SSD-1B.safetensors?download=true)
- 推荐SDXL检查点：[AlbedoBase-XL](https://civitai.com/models/140737)、[Jib-Mix-Realistic-XL](https://civitai.com/models/194768)、[Crystal-Clear-XL](https://civitai.com/models/122822)
- 推荐SD1.5检查点：[Haveall](https://civitai.com/models/213692)
下载完成后将Safetensors格式文件分别放置至`ComfyUImodelscheckpointsSDXL`与`ComfyUImodelscheckpointsSD15`文件夹
11. 注意事项：若从本数据集加载PNG格式工作流，需先点击ComfyUI菜单中的「刷新」按钮，以同步本地已安装的模型，随后按需选择对应的检查点、VAE、LoRA、超分模型、ControlNet模型等资源。

> 「推荐分辨率计算器」（Recommended Resolution Calculator）自定义节点可简化最终分辨率的设定流程，其会自动计算SD模型的初始生成尺寸与超分（或反超分）参数，仅需输入最终目标分辨率即可。

你仅需输入最终目标分辨率即可使用该自定义节点。

# 工作流选择：

将下方PNG图像文件拖拽至ComfyUI界面，即可加载为工作流模板

## 推荐！高效简洁版——单SDXL检查点工作流（LCM-Turbo、提示词风格器、超分模型切换节点、ControlNet、面部细节修复节点）：

（ControlNet参考图像示例：[halo.jpg](https://huggingface.co/datasets/marhensa/comfyui-workflow/resolve/main/halo.jpg)）

## 简洁版——单SD 1.5检查点工作流（LCM、提示词风格器、超分模型切换节点、ControlNet、面部细节修复节点）：

（ControlNet参考图像示例：[halo.jpg](https://huggingface.co/datasets/marhensa/comfyui-workflow/resolve/main/halo.jpg)）

## 拆解版——SDXL工作流（LCM-Turbo、提示词风格器、超分模型切换节点）：

## 拆解版——SDXL-SSD1B工作流（LCM、提示词风格器、超分模型切换节点）：

## 拆解版——SD1.5工作流（LCM、提示词风格器、超分模型切换节点）：

## 极简版——无需额外安装——SDXL/SD1.5 LCM工作流：

原始信息汇总

数据集概述

数据集内容

• 模板文件: 包含Template.JSON和Template.PNG，用于在ComfyUI中加载工作流程模板。
• 工作流程图像: 多种工作流程模板图像，如SDXL-Tidy-LCMTurbo-PromptStyler-UpscaleModelSwitch-ControlNet-FaceDetailer.png、SD15-Tidy-LCM-PromptStyler-UpscaleModelSwitch-ControlNet-FaceDetailer.png等。

安装与使用

• 预安装要求:
  1. 安装"ComfyUI Manager"。
  2. 安装多个自定义节点，包括"ComfyUI-Impact-Pack"、"SDXL Prompt Sytler"等。
  3. 安装"Recommended SD 1.5 Resolution Calculator"。
  4. 安装"ComfyUI ControlNet Aux"及下载相关模型。
  5. 下载并安装LCM LoRA模型。
  6. 下载并安装LCM-Turbo模型。
  7. 下载并安装VAE模型。
  8. 下载并安装Upscale模型。
  9. 下载并安装Checkpoint模型。
• 使用方法:
  1. 拖放Template.JSON或Template.PNG到ComfyUI浏览器界面以加载模板。
  2. 选择并加载所需的工作流程模板图像。

工作流程选项

• 推荐工作流程:
  • Single SDXL Checkpoint Workflow: 使用LCM-Turbo、PromptStyler等工具。
  • Single SD 1.5 Checkpoint Workflow: 使用LCM、PromptStyler等工具。
• 其他工作流程:
  • Breakdown - SDXL Workflow: 使用LCM-Turbo、PromptStyler等工具。
  • Breakdown - SDXL-SSD1B Workflow: 使用LCM、PromptStyler等工具。
  • Breakdown - SD1.5 Workflow: 使用LCM、PromptStyler等工具。
  • Simple without install anything - SDXL / SD15 LCM Workflow: 简化版工作流程。

注意事项

• 在加载PNG工作流程前，需在ComfyUI菜单中点击刷新以更新模型。
• 根据需要选择相应的Checkpoint、VAE、LoRA、Upscale Model和ControlNet模型。